Ultrasonic Guided Waves

Ultrasonic Guided
Waves
Edited by
Cliff J. Lissenden
Printed Edition of the Special Issue Published in Applied Sciences
www.mdpi.com/journal/applsci
Ultrasonic Guided Waves
Ultrasonic Guided Waves
Special Issue Editor

Cliff Lissenden
MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin
Special Issue Editor
Cliff Lissenden
University Park
USA
Editorial Office
MDPI
St. Alban-Anlage 66
4052 Basel, Switzerland
This is a reprint of articles from the Special Issue published online in the open access journal
Applied Sciences (ISSN 2076-3417) (available at: https://www.mdpi.com/journal/applsci/special
issues/UGW).
For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:
LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. Journal Name Year, Article Number,
Page Range.
ISBN 978-3-03928-298-2 (Pbk)

ISBN 978-3-03928-299-9 (PDF)

c 2020 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license, which allows users to download, copy and build upon
published articles, as long as the author and publisher are properly credited, which ensures maximum
dissemination and a wider impact of our publications.
The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons
license CC BY-NC-ND.
Contents
About the Special Issue Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Clifford J. Lissenden
Applied Sciences Special Issue: Ultrasonic Guided Waves
Reprinted from: Appl. Sci. 2019, 9, 3869, doi:10.3390/app9183869 . . . . . . . . . . . . . . . . . . . 1
Christopher Hakoda, Cliff J. Lissenden and Parisa Shokouhi

Clamping Resonators for Low-Frequency S0 Lamb Wave Reflection
Joseph L. Rose, Jason Philtron, Guigen Liu, Yupeng Zhu and Ming Han
A Hybrid Ultrasonic Guided Wave-Fiber Optic System for Flaw Detection in Pipe
Seyed Kamran Pedram, Peter Mudge and Tat-Hean Gan

Enhancement of Ultrasonic Guided Wave Signals Using a Split-Spectrum Processing Method
Houman Nakhli Mahal, Kai Yang and Asoke K. Nandi

Detection of Defects Using Spatial Variances of Guided-Wave Modes in Testing of Pipes

Improved Defect Detection Using Adaptive Leaky NLMS Filter in Guided-Wave Testing
of Pipelines

Defect Detection using Power Spectrum of Torsional Waves in Guided-Wave Inspection
of Pipelines
Qi Wu and Chang-Myung Lee

A Modified Leakage Localization Method Using Multilayer Perceptron Neural Networks in a
Pressurized Gas Pipe
Christopher Hakoda and Cliff J. Lissenden

Using the Partial Wave Method for Wave Structure Calculation and the Conceptual
Interpretation of Elastodynamic Guided Waves
Guopeng Fan, Haiyan Zhang, Hui Zhang, Wenfa Zhu and Xiaodong Chai
Lamb Wave Local Wavenumber Approach for Characterizing Flat Bottom Defects in an
Isotropic Thin Plate
Jianjun Wu, Zhifeng Tang, Keji Yang, Shiwei Wu and Fuzai Lv

Ultrasonic Guided Wave-Based Circumferential Scanning of Plates Using a Synthetic Aperture
Focusing Technique
v
Jianjun Wu, Zhifeng Tang, Keji Yang and Fuzai Lv
Signal Strength Enhancement of Magnetostrictive Patch Transducers for Guided Wave
Inspection by Magnetic Circuit Optimization
Borja Hernandez Crespo, Charles R. P. Courtney and Bhavin Engineer

Calculation of Guided Wave Dispersion Characteristics Using a Three-Transducer
Measurement System
Magdalena Rucka, Erwin Wojtczak and Jacek Lachowicz

Damage Imaging in Lamb Wave-Based Inspection of Adhesive Joints
Shigeaki Okumura, Vu-Hieu Nguyen, Hirofumi Taki, Guillaume Haı̈at, Salah Naili,
and Toru Sato
Rapid High-Resolution Wavenumber Extraction from Ultrasonic Guided Waves Using
Adaptive Array Signal Processing
Bin Wang, Yihui Da and Zhenghua Qian

Forward and Inverse Studies on Scattering of Rayleigh Wave at Surface Flaws
Lei Zhang, Yuan Yang, Xiaoyuan Wei and Wenqing Yao

The Study of Non-Detection Zones in Conventional Long-Distance Ultrasonic Guided Wave
Inspection on Square Steel Bars
Bo Xing, Zujun Yu, Xining Xu, Liqiang Zhu and Hongmei Shi
Research on a Rail Defect Location Method Based on a Single Mode Extraction Algorithm
Pengfei Zhang, Zhifeng Tang, Fuzai Lv and Keji Yang

Numerical and Experimental Investigation of Guided Wave Propagation in a Multi-Wire Cable
Seung Jin Chang and Seung-Il Moon

Compensation for Group Velocity of Polychromatic Wave Measurement in Dispersive Medium
Yevgeniya Lugovtsova, Jannis Bulling, Christian Boller and Jens Prager

Analysis of Guided Wave Propagation in a Multi-Layered Structure in View of Structural
Health Monitoring
vi
About the Special Issue Editor
Cliff Lissenden is a professor of engineering science and mechanics at Penn State. He joined
the Department of Engineering Science and Mechanics in 1995 and gained a joint appointment
in Acoustics in 2011. He is an ASME Fellow and was the founder and director of the Ben
Franklin Center of Excellence in Structural Health Monitoring. His current research is mainly
focused on the nondestructive characterization of materials using ultrasonic guided waves and is
broadly applicable to metals, composites, concrete, rock, and bone. Characterization systems use
piezoelectric, magnetostrictive, electromagnetic, and laser-based transduction of ultrasonic waves
for nondestructive testing, inspection, and monitoring, often in harsh environments. The basic
physics employed are that the wave speeds depend on the elastic properties, that acoustic
impedance mismatches cause wave scattering, and that material nonlinearity distorts the waveform.
Current applications include robotic nondestructive inspection of stress corrosion cracking, structural
integrity analysis of bonded joints in composites, detection of incipient damage with nonlinear
ultrasonics, process monitoring of additive manufacturing with laser ultrasonics, cloaking critical
infrastructure from earthquakes, and characterization of bone healing.
vii
applied
sciences
Editorial
Applied Sciences Special Issue: Ultrasonic
Guided Waves
Clifford J. Lissenden
Department of Engineering Science and Mechanics, Pennsylvania State University, University Park, PA 16802,
USA; lissenden@psu.edu
Received: 9 September 2019; Accepted: 10 September 2019; Published: 15 September 2019
The propagation of ultrasonic guided waves in solids is an important area of scientific inquiry
due primarily to their practical applications for the nondestructive characterization of materials, such
as nondestructive inspection, quality assurance testing, structural health monitoring, and for achieving
material state awareness. This Special Issue of the journal covers all aspects of ultrasonic guided waves
(e.g., phased array transducers, meta-materials to control wave propagation characteristics, scattering,
attenuation, and signal processing techniques) from the perspective of modeling, simulation, laboratory
experiments, or field testing. In order to fully utilize ultrasonic guided waves for these applications, it
is necessary to have a firm grasp of their requisite characteristics, which include being multimodal,
dispersive, and comprised of unique displacement profiles through the thickness of the waveguide.
The majority of the manuscripts in this Special Issue report on waveguides that are pipes and plates,
but rods, rail, multi-strand wire, human bone, and half-spaces are also investigated. Hakoda et al. [1]
provide experimental results on an aluminum plate, demonstrating how wave propagation can be
manipulated by changing the boundary conditions. Specifically, surface-mounted resonators block
the propagation of the S0 Lamb mode at low frequencies. Forbidding the propagation of the S0
Lamb mode turns out to be very different from forbidding the propagation of the A0 Lamb mode
but can be explained in terms of the boundary conditions, which provide an alternative approach to
meta-materials. The remaining 18 manuscripts pertain directly to the detection of defects within the
waveguide in order to enable the assessment of the structural integrity.
While the purpose of pipes is to transport or circulate fluids, their hollow cylindrical geometry
forms an efficient waveguide for longitudinal, torsional, and flexural ultrasonic waves. Torsional
waves are often preferred for long-range defect detection because the wave energy does not leak into
the fluids inside the pipe. Six manuscripts in the Special Issue investigate ultrasonic guided waves
in pipes for the purpose of defect detection. Rose et al. [2] show that an array of fiber Bragg grating
(FBG) sensors multiplexed along an optical fiber can receive torsional guided waves in a pipe, thus
improving the range of piezoelectric and magnetostrictive actuators because reflections do not have to
propagate back to the actuator. Pedram et al. [3] demonstrate the use of split-spectrum processing
to improve the signal-to-noise ratio (SNR) of axisymmetric torsional waves in pipes and investigate
its limitations. They show a 30 dB increase in SNR. Mahal et al. [4] first propose a method to detect
defects in pipes from axisymmetric torsional waves having a low signal-to-noise ratio by leveraging
the signals received from individual elements in the ring array transducer. Then, they [5] propose an
adaptive leaky normalized least-mean-square filter to reduce the effect of unwanted flexural waves on
the detection of axisymmetric waves. In a third paper [6], the same authors describe the exploitation
of the power spectrum to distinguish torsional waves from flexural waves in pipes for more reliable
defect detection. Taking a different approach, Wu and Lee [7] propose applying a multilayer perceptron
neural network to detect leaks in pressurized natural gas pipelines, based upon acoustic emissions
from the leaks. Accurate localization depends on the group velocity of the flexural waves generated
during the acoustic emission events.
Appl. Sci. 2019, 9, 3869; doi:10.3390/app9183869 1 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, 3869
The Lamb waves and shear-horizontal (SH) waves that propagate in traction-free homogeneous
isotropic plates are often used for defect detection when access is limited or to cover a large area.
Hakoda and Lissenden [8] demonstrate how the analysis of partial waves provides new perspectives
on the propagation characteristics of elastodynamic guided waves in plates as well as other types of
waveguides. Fan et al. [9] employ wavenumber filtering to remove Lamb modes other than the A0
mode to detect flat bottom holes in a plate. Wu et al. [10] propose a methodology for nondestructive
inspection of cylindrical storage tanks’ base plates whereby a transducer is scanned around the
perimeter of the base plate. The generated fundamental SH waves propagate in the radial direction of
the base plate. Defects scatter the waves, and reflections are received by the transducer. A synthetic
aperture focusing technique based on the so-called exploding reflector model is applied to image the
tank bottom. Wu et al. [11] then demonstrate a 12 dB improvement in the amplitude of SH waves
generated by a magnetostrictive transducer when a soft magnetic patch is applied, which reduces
the resistance of the magnetic circuit. This optimization of the magnetic circuit is also applicable to
transducers that send and receive torsional waves in pipes. Crespo et al. [12] propose an experimental
method to construct the Lamb wave and SH wave dispersion curves for a plate using a transmitter and
two receivers, which requires no knowledge of the material properties. The paper by Rucka et al. [13]
completes the applications to plates by investigating the integrity of the adhesive in single-lap joints
between aluminum plates by images processed from the wavefield acquired by scanning laser Doppler
vibrometer and the weighted root mean square.
Axial transmission of ultrasound in human cortical bone is under development as a biomarker for
early detection of osteoporosis and fracture risk. While the geometry of human bones, such as the
tibia, is far from the geometry of plates, it is common to use Lamb waves as surrogates for waves
in cortical bones. Okumura et al. [14] developed a rapid high-resolution technique to determine the
wavenumbers of propagating waves in surrogates for human bone.
One manuscript addresses Rayleigh waves propagating along the surface of a half-space.
Wang et al. [15] provide forward and inverse analyses of Rayleigh wave scattering from surface breaking
defects based on reflection coefficients in order to predict the shape of realistic surface-breaking flaws.
The final four manuscripts investigate defect localization for long-range inspection of different
types of one-dimensional waveguides. Zhang et al. [16] considered long-range ultrasonic guided
waves in a steel square bar and analyzed the non-detection zone for flaws located near the back
wall. Xing et al. [17] propose a single mode extraction algorithm to precisely localize defects in rail.
Zhang et al. [18] analyze guided waves, including second harmonic generation associated with the
contact acoustic nonlinearity, in multi-strand cable where transducers are coupled to only two strands
in the cable. Finally, Chang and Moon [19] investigate accurate group velocity of electromagnetic
waves in dispersive waveguides, such as insulated cables, in which defect localization is performed
using reflectometry.
Funding: This research received no external funding.

Acknowledgments: I would like to thank all of the authors who submitted manuscripts to this special issue, as
well as the reviewers, who helped to improve the quality of the manuscripts.
Conflicts of Interest: The author declares no conflicts of interest.
References
1. Hakoda, C.; Lissenden, C.; Shokouhi, P. Clamping resonators for low-frequency S0 lamb wave reflection.
Appl. Sci. 2019, 9, 257. [CrossRef]
2. Rose, J.; Philtron, J.; Liu, G.; Zhu, Y.; Han, M. A hybrid ultrasonic guided wave-fiber optic system for flaw
detection in pipe. Appl. Sci. 2018, 8, 727. [CrossRef]
3. Pedram, S.; Mudge, P.; Gan, T. Enhancement of ultrasonic guided wave signals using a split-spectrum
processing method. Appl. Sci. 2018, 8, 1815. [CrossRef]
2
Appl. Sci. 2019, 9, 3869
4. Mahal, H.; Yang, K.; Nandi, A. Detection of defects using spatial variances of guided-wave modes in testing
of pipes. Appl. Sci. 2018, 8, 2378. [CrossRef]
5. Nakhli Mahal, H.; Yang, K.; Nandi, A. Improved defect detection using adaptive leaky NLMS filter in
guided-wave testing of pipelines. Appl. Sci. 2019, 9, 294. [CrossRef]
6. Nakhli Mahal, H.; Yang, K.; Nandi, A. Defect detection using power spectrum of torsional waves in
guided-wave inspection of pipelines. Appl. Sci. 2019, 9, 1449. [CrossRef]
7. Wu, Q.; Lee, C. A modified leakage localization method using multilayer perceptron neural networks in a
pressurized gas pipe. Appl. Sci. 2019, 9, 1954. [CrossRef]
8. Hakoda, C.; Lissenden, C. Using the partial wave method for wave structure calculation and the conceptual
interpretation of elastodynamic guided waves. Appl. Sci. 2018, 8, 966. [CrossRef]
9. Fan, G.; Zhang, H.; Zhang, H.; Zhu, W.; Chai, X. Lamb wave local wavenumber approach for characterizing
flat bottom defects in an isotropic thin plate. Appl. Sci. 2018, 8, 1600. [CrossRef]
10. Wu, J.; Tang, Z.; Yang, K.; Wu, S.; Lv, F. Ultrasonic guided wave-based circumferential scanning of plates
using a synthetic aperture focusing technique. Appl. Sci. 2018, 8, 1315. [CrossRef]
11. Wu, J.; Tang, Z.; Yang, K.; Lv, F. Signal strength enhancement of magnetostrictive patch transducers for
guided wave inspection by magnetic circuit optimization. Appl. Sci. 2019, 9, 1477. [CrossRef]
12. Hernandez Crespo, B.; Courtney, C.; Engineer, B. Calculation of guided wave dispersion characteristics using
a three-transducer measurement system. Appl. Sci. 2018, 8, 1253. [CrossRef]
13. Rucka, M.; Wojtczak, E.; Lachowicz, J. Damage imaging in lamb wave-based inspection of adhesive joints.
14. Okumura, S.; Nguyen, V.; Taki, H.; Haïat, G.; Naili, S.; Sato, T. Rapid high-resolution wavenumber extraction
from ultrasonic guided waves using adaptive array signal processing. Appl. Sci. 2018, 8, 652. [CrossRef]
15. Wang, B.; Da, Y.; Qian, Z. Forward and inverse studies on scattering of rayleigh wave at surface flaws. Appl.
Sci. 2018, 8, 427. [CrossRef]
16. Zhang, L.; Yang, Y.; Wei, X.; Yao, W. The study of non-detection zones in conventional long-distance ultrasonic
guided wave inspection on square steel bars. Appl. Sci. 2018, 8, 129. [CrossRef]
17. Xing, B.; Yu, Z.; Xu, X.; Zhu, L.; Shi, H. Research on a rail defect location method based on a single mode
extraction algorithm. Appl. Sci. 2019, 9, 1107. [CrossRef]
18. Zhang, P.; Tang, Z.; Lv, F.; Yang, K. Numerical and experimental investigation of guided wave propagation
in a multi-wire cable. Appl. Sci. 2019, 9, 1028. [CrossRef]
19. Chang, S.; Moon, S. Compensation for group velocity of polychromatic wave measurement in dispersive
medium. Appl. Sci. 2017, 7, 1306. [CrossRef]
© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
3
applied
sciences
Article
Clamping Resonators for Low-Frequency S0 Lamb
Wave Reflection
Christopher Hakoda, Cliff J. Lissenden * and Parisa Shokouhi
Department of Engineering Science and Mechanics, Pennsylvania State University, University Park,
Pennsylvania 16802, PA, USA; cnh137@psu.edu (C.H.); parisa@engr.psu.edu (P.S.)
* Correspondence: lissenden@psu.edu
Received: 30 November 2018; Accepted: 9 January 2019; Published: 12 January 2019
Abstract: A recent elastic metamaterial study found that resonators that “clamp” a plate waveguide
can be used to create a frequency stop-band gap. The result was that the resonator array can prohibit
the propagation of an A0 Lamb wave mode. This study investigates whether the concept can be
extended to S0 Lamb wave modes by designing resonators that can prohibit the propagation of
S0 Lamb wave modes in a 1-mm aluminum plate waveguide at 50 kHz. The frequency-matched
resonators did not reduce the transmitted signal, leading to the conclusion that the design concept of
frequency-matched resonators is not always effective. On the other hand, the resonators designed
to clamp the upper surface of the plate were very effective and reduced the transmitted signal by
approximately 75%.
Keywords: metamaterial; resonator; low-frequency; lamb wave
1. Introduction
The field of metamaterials has emerged over the past 20 years and inspired numerous innovations
in optics, electromagnetics e.g., [1–3] and more recently in acoustics e.g., [4–6]. Phononic crystals
and acoustic metamaterials are composites engineered to mold elastic wave dispersion through
scattering or local resonances to achieve desired spectral and phase properties such as band gaps
around specific frequencies. Phononic crystals are composed of a periodic array of scatterers.
The mechanism for the formation of band gaps in phononic crystals is a Bragg-like scattering of
waves with wavelength comparable to the dimension of the period of the crystal or the crystal
lattice constant [7]. This characteristic limits the use of phononic crystals as low-frequency filters.
However, low-frequency filters can be achieved by using a different class of acoustic metamaterials
that comprise a series of locally resonating units which are not necessarily periodic. Liu et al. [8]
demonstrated that such assemblies are capable of muting waves of wavelengths that are orders of
magnitude longer than the lattice constant. By tuning the properties of the individual units and their
arrangement, the metamaterial can act as a low-frequency band-gap filter (blocking the incoming
elastic energy within certain frequency bands) or wedge (converting the mode and diverting the
incident wave). The studies of resonators that followed have been mostly parametric in nature and
focused on designing elements with strong frequency-coupled characteristics that will yield band gaps
about a given center frequency [9–11].
As noted above, the field of elastodynamic metamaterials is relatively new, but nonetheless
presents a wide variety of resonator design methodologies [9–13]. An experiment conducted by
Rupin et al. [14] found that cylindrical rods adhered to a plate waveguide could be used to prohibit
the propagation of the A0 Lamb wave mode at low frequencies (<11 kHz). Soon thereafter, a paper by
Williams et al. [15] sought to determine what characteristic of the cylindrical rods contributed to the
A0 Lamb wave mode band gap. They found that the rods vibrated in an axial mode which applied

Appl. Sci. 2019, 9, 257
a “clamping”-like effect to the waveguide. That is, the rods would vibrate in such a way that the
out-of-plane displacement at their bases was minimized. They also compared the frequencies at which
the minimal displacements occurred with the band gaps reported in Rupin et al.’s paper and found a
good agreement between the two spectra.
The spectrum detailed in Williams et al.’s paper was calculated using a vibration model subjected
to a harmonic forcing function. Upon further investigation, we found that the frequencies at which
minimum displacement occurred were equal to the resonant frequencies of an axially vibrating,
fixed-free beam. Since these resonant frequencies have well-known solutions, this finding presented
itself as a very straightforward method for designing sub-wavelength resonators for the A0 Lamb
wave mode to create band gaps around specific frequencies. However, this approach was only
shown to be effective for low-frequency A0 Lamb wave modes which have primarily out-of-plane
displacement. The purpose of this article is to extend this concept to design resonators that prohibit
the propagation of low-frequency S0 Lamb wave modes and to determine its applicability at higher
frequencies. To accomplish this, two resonator designs, one designed based on a frequency matching
approach and the other based on the sole concept of clamping, were tested to see how well they can
impede the propagation of a 50 kHz S0 Lamb wave. These resonators are hereafter referred to as
the frequency-matched resonators and the clamping resonators, respectively. While the design of
the frequency-matched resonators is straightforward (based on resonance), the theoretical basis for
the design of the clamping resonators is more involved and requires the full attention of a separate
publication. We believe that a rational design protocol is possible and will use the experimental results
provided herein to demonstrate it in a subsequent publication. Thus, in this article we simply provide
finite-element simulations to qualitatively confirm the experimental findings.
2. Experimental Setup
For the experiment, the consistency of various experimental parameters was prioritized to
obtain a comparison between the resonator types. That is, various experimental parameters are
the same regardless of the resonators being tested. This includes the transducer positions, waveguide
dimensions, resonator spacing/positioning, the area of contact between the resonator and plate,
resonator material properties, and overall resonator mass. The only difference between tests is the
shape of the resonators.
The guided waves that will be used for the test are the S0 Lamb wave mode with a 50 kHz 5-cycle
pulse, which corresponds with a wavenumber of 58.1 rad/m. At these low frequencies, this mode
corresponds with a wave-structure (i.e., displacement profile) that is effectively planar and consists of
in-plane displacement through the thickness of the plate waveguide. The S0 Lamb wave mode was
excited and received using shear transducers with the polarization aligned with the wave-propagation
direction. A schematic of the setup, including transducer placement and resonator positions, is shown
in Figure 1. Three rows of ten resonators were used for the resonator array and adhered to the surface
of the waveguide using cyanoacrylate adhesive. The spacing of the resonator array shown in Figure 1b
is based on how close together the clamping resonators could be positioned.
The main method of assessing the resonators will be to compare two-dimensional fast Fourier
transforms (2DFFTs) of the signal transmitted past the resonator array [16]. The receiving transducer
was moved away from the resonator array in 2 cm increments. This was done 20 times to measure an
A-scan at 20 positions along the expected wave-propagation direction. For the 2DFFT, zero-padding
was applied to increase the number of bins in the wavenumber spectrum, but no zero-padding was
applied for the frequency spectrum.
6
Appl. Sci. 2019, 9, 257
(a) (b)
Figure 1. Schematic (not to scale) of experimental setup: (a) top-view of the transducer positions with
respect to the resonator array, and (b) a zoomed in view of the resonator array that shows how the
resonators are positioned with respect the wave-propagation direction. The squares are the contact
areas between the resonators and the plate waveguide.
3. Frequency-matched Resonators
Analogous to Rupin et al.’s [14] axial rod design, the frequency-matched resonator was designed
to vibrate at 50 kHz in a flexural mode under fixed-free boundary conditions. The fixed boundary
condition was expected to apply the desired clamping effect described in Williams et al. [15].
The refined design of these frequency-matched resonators consists of a 7.9 × 7.9 mm square rod
that is 23.4 mm long. Figure 2a shows the 2DFFT of a baseline measurement which was taken prior
to bonding the resonators to the plate waveguide. The peak at a wavenumber of −58.1 rad/m and a
frequency of 53.9 kHz shown in Figure 2a indicates that the S0 mode is being excited.
The Frequency-matched Resonators were positioned on the waveguide as shown in Figure 1
and adhered to the plate waveguide. The A-scans described in section 2 were measured and used to
calculate the 2DFFT shown in Figure 2b.
(a) (b) (c)
Figure 2. The 2DFFT of the transmission measurements for (a) the Baseline, (b) the Frequency-matched
Resonator and (c) the Clamping Resonator. The diagonal lines are the dispersion curves for the forward
and backward propagating S0 Lamb wave modes. The vertical and horizontal lines show where the
frequency- and wavenumber-spectrum slices shown in Figure 6 are obtained. The peak values are
noted in each figure.
When comparing Figure 2a to Figure 2b, the frequency-matched resonators do not reduce
the transmitted S0 signal. This is believed to be due to the resonators vibrating differently than
expected. That is, unlike the axial vibration described in Williams et al., the flexural vibration of the
frequency-matched resonator is likely to cause unwanted out-of-plane displacement.
The frequency-matched resonators are an attempt at designing resonators based on resonant
frequency, but as shown, it was not an effective approach for this application. In the next section,
we propose a clamping resonator that was designed to effectively clamp the S0 wave’s in-plane motion
without introducing unwanted out-of-plane displacements. That is, we focused on how the resonator
clamps the surface of the waveguide.
7
Appl. Sci. 2019, 9, 257
4. Clamping Resonators
The clamping resonators (shown in Figure 3) were designed using symmetry as a means to
reduce the unwanted displacement components. The resonators are composed of a center cube with
a side length of 7.9 mm, and four cylindrical “arms” that are 4.8 mm in diameter and 15 mm long.
These components are joined using cyanoacrylate adhesive so that the resonator shown in Figure 3
is built. The arms perpendicular to the wave-propagation direction are meant to clamp the in-plane
displacement, while the arms parallel to the wave-propagation direction are meant to reduce the
unwanted out-of-plane displacement. All these arms undergo flexural vibration, presumably at
the same frequency as the propagating wave, to achieve this effect. The guiding principles of the
conceptual design were achieved through a combination of theoretical modeling and finite-element
analysis, which will be described in more detail in a subsequent publication.
(a) (b)
Figure 3. (a) Isometric view of a clamping resonator, and (b) the 30 clamping resonators in an array
according to the spacing shown in Figure 1b.
4.1. Transmission beyond the Clamping Resonator Array

A cursory look at Figure 2c reveals that compared to the baseline results shown in Figure 2a,
the clamping resonators noticeably reduce the amplitude of the transmitted S0 Lamb wave mode.
By extracting slices of data from Figure 2 that are parallel to the frequency or wavenumber axis we can
plot the frequency spectrum, or the wavenumber spectrum as shown in Figure 4a and b, respectively.
These spectra reveal that the clamping resonator reduces the S0 Lamb wave mode by about 75% of the
baseline measurement. On the other hand, the frequency-matched resonator unexpectedly causes a
very slight increase in the transmitted S0 mode.
(a) (b)
Figure 4. (a) The frequency spectra of the transmitted signal, which were extracted from the 2DFFTs
shown in Figure 2 at a wavenumber of −58.1 rad/m; (b) The wavenumber spectra of the transmitted
signal, which were extracted from the 2DFFTs shown Figure 2 at a frequency of 53.9 kHz.
8
Appl. Sci. 2019, 9, 257
Examples of the transmitted waveforms for the baseline and clamping resonators are shown
in Figure 5. The comparison shows that the wave packet received beyond the clamping resonators
has a significantly reduced amplitude at its center (around 200 μs). The inclusion of the clamping
resonators caused the measured signals to vary slightly more than the signals from the baseline
and frequency-matched resonator measurements; presumably due to scattering from the resonators.
We note that putty was used to reduce edge reflections, but it did not completely attenuate the wayward
waves, which could explain why some noise remains in the A-scans shown in Figure 5. The 2DFFT
was used to help reduce unwanted signals reflecting off the edges of the plate by organizing the
components of the A-scan measurements by wavenumber and frequency.
Figure 5. A-scans measured 69 cm from the transmitting transducer. The A-scan (black) from the plate
with clamping resonators has a reduced amplitude around 200 μs when compared to the baseline
(grey). These A-scan examples were some of the noisier signals that were measured.
4.2. Reflection from the Clamping Resonator Array

To determine whether the clamping resonators reflect or attenuate the incident S0 Lamb wave,
the transducer setup was changed to measure and calculate the 2DFFT of possible reflections as shown
in Figure 6a. The incident/reflection A-scans were used to calculate the 2DFFT shown in Figure 6b.
Extracting a slice of the data parallel to the wavenumber axis at a frequency of 53.9 kHz yields the
wavenumber spectrum shown in Figure 6c. The wavenumber spectrum indicates that about 62% of
the incident wave was reflected from the resonator array. It should be noted that the plot shown in
Figure 6c is not directly comparable to those in Figure 4b because different transducer arrangements
were leveraged.
9
Appl. Sci. 2019, 9, 257
Figure 6. (a) Schematic of transducer positioning with respect to the resonator array for measuring
reflected waves; (b) The 2DFFT of the incident/reflection measurements. The diagonal blue lines are
the dispersion curves for the forward and backward propagating S0 Lamb wave. The vertical black
line is where the wavenumber-spectrum slice data at a frequency of 53.9 kHz are obtained for (c) the
wavenumber spectrum.
4.3. Finite-Element Simulations

Frequency domain finite-element analysis [17] was performed in order to provide confidence
in the experimental results presented in Figures 2 and 4. This simulation assumes a plane wave
solution and uses the frequency domain by assuming eiωt dependence, hence for a forward traveling
wave the wavenumber k x is negative-valued. Perfect matching layers are used on the ends of the
model to prevent “end-wall” reflections. The remaining parts of the domain are divided into buffer,
excitation, incident/transmit, and “affected” regions. The purpose of this model is to evaluate the
transmission and reflection qualities of reflectors defined in the “affected” region. A similar model was
used in Hsu [18], but we apply a body-load excitation instead of line excitation to excite a single mode.
The displacement component u x is plotted along the x-axis in Figure 7. The y-direction width of the
model was chosen to correspond with the spacing of the resonators, which equates to a periodically
repeating array of resonators in the ±y-directions since periodic boundary conditions are applied to
these edges. The clamping resonators are effective at prohibiting the propagation of much of the S0
Lamb wave, while the frequency-matched resonators are completely ineffective. The wavenumber
spectrum was calculated using the complex-displacements measured from the transmitted wavefield
and is shown in Figure 8. At the wavenumber of the S0 mode we see that the frequency-matched
resonators only reduce the spectrum by approximately 7%, while the clamping resonators reduce it
by 90%.
10
Appl. Sci. 2019, 9, 257
Figure 7. Finite-element predictions of the x-component displacement of an S0 Lamb wave at 50 kHz

plotted along the x-axis showing that the clamping resonators block wave propagation, while the
frequency-matched resonators do not.
Figure 8. Finite-element-predicted wavenumber spectrum of the guided waves that were allowed
to transmit past the resonator array. This spectrum is expected to only contain forward (+x)
traveling waves.
5. Conclusions
Based on Williams et al.’s [7] identification of the clamping phenomenon, we sought to investigate
whether the concept’s applicability could be extended to S0 Lamb wave modes. Two approaches for
designing a resonator were used based on the concepts discussed by Williams et al. These approaches
resulted in the frequency-matched resonators and the clamping resonators (based on the clamping
concept). The clamping resonators performed well by decreasing the amplitude of the transmitted
11
Appl. Sci. 2019, 9, 257
wave by about 75%. When measuring the reflection from the clamping resonators, 62% of the incident
wave was found to be reflected. The reason these two percentages do not equal each other is
provisionally attributed to scattering from the resonator array and beam spread. In comparison,
the frequency-matched resonators did not reduce the S0 Lamb wave amplitude. While the results
for the clamping resonators used in these experiments are limited to the S0 Lamb mode at 50 kHz,
these results and those of Williams et al. suggest that by designing a resonator based on the concept of
clamping the waveguide, a metamaterial capable of prohibiting the propagation of a low-frequency A0
or S0 Lamb wave is possible. This also suggests that these resonators could be mode-specific by using
different resonator designs for each mode. Moreover, a more detailed description of the basis for the
concepts used to design the clamping resonators will be provided along with the associated theoretical
modeling effort, which is still ongoing. Future work will focus on the frequency stop-bandwidth.
Author Contributions: Conceptualization, P.S., C.L. and C.H.; investigation, C.H.; writing—original draft
preparation, C.H.; writing—review and editing, C.H., C.L. and P.S.; supervision, C.L.; project administration, P.S.;
funding acquisition, P.S.
Funding: This research was funded by a seed grant from the College of Engineering at The Pennsylvania
State University, “Engineering a giant metamaterial: a band-stop seismic/blast filter to shield critical civil
infrastructure”.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Pendry, J.B. Negative refraction makes a perfect lens. Phys. Rev. 2000, 85, 3966–3969. [CrossRef] [PubMed]
2. Cubukcu, E.; Aydin, K.; Ozbay, E.; Foteinopoulou, S.; Soukoulis, C.M. Negative refraction by photonic
crystals. Nature 2003, 423, 604–606. [CrossRef] [PubMed]
3. Schurig, D.; Mock, J.J.; Justice, B.J.; Cummer, S.A.; Pendry, J.B.; Starr, A.F.; Smith, D.R. Metamaterial
electromagnetic cloak at microwave frequencies. Science 2006, 314, 977–980. [CrossRef] [PubMed]
4. Norris, A.N.; Shuvalov, A.L. Elastic cloaking theory. Wave Motion 2011, 48, 525–538. [CrossRef]
5. Farhat, M.; Guenneau, S.; Enoch, S. Ultrabroadband elastic cloaking in thin plates. Phys. Rev. Lett. 2009, 103,
1–4. [CrossRef] [PubMed]
6. Stenger, N.; Wilhelm, M.; Wegener, M. Experiments on elastic cloaking in thin plates. Phys. Rev. Lett. 2012,
108, 1–5. [CrossRef] [PubMed]
7. Deymier, P.A. (Ed.) Acoustic Metamaterials and Phononic Crystals; Springer: Berlin, Germany, 2009.
8. Liu, Z.; Zhang, X.; Mao, Y.; Zhu, Y.Y.; Yang, Z.; Chan, C.T.; Sheng, P. Locally resonant sonic materials. Science
2000, 289, 1734–1736. [CrossRef] [PubMed]
9. Bilal, O.R.; Hussein, M.I. Trampoline metamaterial: Local resonance enhancement by springboards.
Appl. Phys. Lett. 2013, 103, 111901. [CrossRef]
10. Wu, T.C.; Wu, T.T.; Hsu, J.C. Waveguiding and frequency selection of lamb waves in a plate with a periodic
stubbed surface. Phys. Rev. B 2009, 79. [CrossRef]
11. Rupin, M.; Roux, P. A multi-wave elastic metamaterial based on degenerate local resonances. J. Acoust. Soc.
Am. Express Lett. 2017, 142, EL75–EL81. [CrossRef] [PubMed]
12. Li, Y.; Zhu, L.; Chen, T. Plate-type elastic metamaterials for low-frequency broadband elastic wave
attenuation. Ultrasonics 2017, 73, 34–42. [CrossRef] [PubMed]
13. Zhao, J.; Bonello, B.; Boyko, O. Focusing of the lowest-order antisymmetric lamb mode behind a
gradient-index acoustic metalens with local resonators. Phys. Rev. B 2016, 93, 174306. [CrossRef]
14. Rupin, M.; Lemoult, F.; Lerosey, G.; Roux, P. Experimental demonstration of ordered and disordered
multiresonant metamaterials for lamb waves. Phys. Rev. Lett. 2014, 112, 234301. [CrossRef] [PubMed]
15. Williams, E.G.; Roux, P.; Rupin, M.; Kuperman, W. Theory of multiresonant metamaterials for A0 lamb
waves. Phys. Rev. B 2015, 91, 104307. [CrossRef]
16. Alleyne, D.N.; Cawley, P. The interaction of lamb waves with defects. IEEE Trans. Ultrason. Ferroelectr. Freq.
Control 1992, 39, 381–397. [CrossRef] [PubMed]
12
Appl. Sci. 2019, 9, 257
17. Chillara, V.K.; Ren, B.; Lissenden, C.J. Guided wave mode selection for inhomogeneous elastic waveguides
using frequency domain finite element approach. Ultrasonics 2016, 67, 199–211. [CrossRef] [PubMed]
18. Hsu, J. Low-frequency forbidden band in phononic crystal plates with Helmholtz resonators. Jpn. J. Appl.
Phys. 2011, 50. [CrossRef]
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
13
applied
sciences
Article
A Hybrid Ultrasonic Guided Wave-Fiber Optic
System for Flaw Detection in Pipe
Joseph L. Rose 1,2 , Jason Philtron 1, *, Guigen Liu 3 , Yupeng Zhu 3 and Ming Han 3
1 FBS, Inc. dba Guidedwave, 450 Rolling Ridge Drive, Bellefonte, PA 16823, USA;
jlresm@engr.psu.edu or jrose@gwultrasonics.com
2 Department of Earth and Engineering Sciences, Pennsylvania State University, University Park,
PA 16802, USA
3 Department of Electrical and Computer Engineering, Michigan State University, East Lansing,
MI 48824-1226, USA; guigenliu@hotmail.com (G.L.); yup.zhu@gmail.com (Y.Z.); mhan@egr.msu.edu (M.H.)
* Correspondence: jphiltron@gwultrasonics.com
Received: 11 April 2018; Accepted: 2 May 2018; Published: 5 May 2018
Abstract: The work presented in this paper shows that Fiber Bragg Grating (FBG) optical fiber sensors
can potentially be used as receivers in a long-range guided wave torsional-mode pipe inspection
system. Benefits over the conventional pulse-echo method arise due to reduced total travel distance
of the ultrasonic guided wave reflections, since reflections from defects and structural features do not
need to propagate a full round trip back to the transmitting collar. This is especially important in
pipe configurations with high attenuation, such as coated and buried pipelines. The use of FBGs as
receivers instead of conventional piezoelectric or magnetostrictive elements also significantly reduces
cabling, since multiple FBG receivers can be placed along a single optical fiber which has a diameter
on the order of only around 100 μm. The basic approach and sample results are presented in the
paper. Additionally, a brief overview of some topics in ultrasonic guided waves is presented as a
background to understand the inspection problem presented here.
Keywords: ultrasonic guided waves; fiber optics; fiber Bragg grating; nondestructive testing;
structural health monitoring; pipe inspection
1. Introduction
Guided wave ultrasound is increasingly being used to solve inspection problems in a wide variety
of applications. This is because guided waves can be used to inspect over long distances from a single
probe position, offer complete volumetric coverage, and can inspect hidden/inaccessible structures,
such as those under water, coatings, insulation, and soil. Before discussing the merger of ultrasonic
guided waves with fiber optic technology, the major topic in this paper, a brief review of some of our
first accomplishments on various topics in ultrasonic guided waves is presented. This illustrates the
background leading to the topic of this paper.
Two textbooks on the subject of wave mechanics fundamentals with emphasis on guided
waves for nondestructive testing (NDT) and structural health monitoring (SHM) are presented by
Rose, [1] 1999, [2] 2014. Some selected developments over the last few decades include the following.
The value of computing and selecting specific wave structures from the phase velocity dispersion
curves was initially pointed out by Ditri, Rose, and Chen [3] 1991, where defects close to a surface
were found with wave structure having lots of energy near the surface. Since then paying attention
to wave structure has led to solutions for many important problems in NDT and SHM. Pilarski,
Ditri, and Rose in [4] 1993, demonstrated the ability to locate wave structure profiles with in-plane
displacement on the surface of a waveguide, which has played a significant role in ice detection and
the impact of water loaded surfaces on many NDT and SHM problems. A variety of problems in

Appl. Sci. 2018, 8, 727
the aircraft industry are covered by Rose and Soley [5] 2000. Such topics as lap joint, tear strap and
honeycomb inspection are discussed. A method of focusing ultrasonic guided waves in pipe was
introduced by Rose, Sun, Mudge, and Avioli [6] 2003. The phased array focusing approach improves
detection sensitivity, axial and lateral resolution and even penetration power. The ability to generate
dispersion curves for structures having an arbitrary cross section was introduced by Hayashi, Song,
and Rose [7] 2003. Many new applications with guided waves became possible and in particular
a breakthrough in treating guided wave propagation in rails. Guided wave tomography concepts
and several sample problems were coved by Gao, Shi, and Rose [8] 2005. The critical problem of rail
inspection under shelling for transverse crack detection in the rail head along with possibilities of
web and base defect detection was presented by Lee, Rose, and Cho [9] 2008. Adhesive bondline
inspection breakthroughs was reported for certain substrate systems by Puthillath and Rose [10] 2010.
Utilization of magnetostrictive technology advancements for the generation and utilization of guided
waves was reported by Van Velsor, Royer, Owens, and Rose [11] 2013. Guided wave mode and
frequency tuning and optimization is discussed in Philtron and Rose [12] 2014, making it possible
to find defects in a variety of different structures more completely. A phased array system for plate
inspection was developed and reported by Rose, Borigo, Owens, and Reese [13] 2017. This technique
is responsible for amazingly fast inspection of corrosion type defects in large plate-like structures
employing radar type circumferential or sector scans. Other NDT applications for guided waves, for
example, include oil-tank weld detection using EMAT-generated shear-horizontal guided waves [14]
2017, fundamental torsional mode detection of angled cut flaws [15] 2016, and liquid level displacement
measurements using magnetostrictive sensors [16] 2017.
Our background in ultrasonic guided waves thus prepared us for a merger with concepts
presented by Perez, Cui, and Udd [17] 2001. Perez et al. [17] showed the possibility of ultrasonic
wave interaction with fiber Bragg gratings (FBG) in an optical fiber. To date, FBG ultrasonic sensors
have been extensively investigated [18] 2008. The use of the fundamental shear horizontal mode in
plates and torsional mode in pipes is of interest for ultrasonic inspection due to the nondispersive
nature and insensitivity to water loading that these modes offer. However, the majority of FBG
ultrasonic sensors have been used to detect Lamb-type guided waves which are quite different from
the shear horizontal (SH) waves in plates and torsional waves in pipes, there are only a few papers
experimentally describing the usage of FBG sensors for SH wave detection [19–21] but with inclusive
results. The mechanism governing the interaction of an FBG sensor with SH waves in plates has
not been thoroughly reported until our work was recently published by Liu, Philtron, Zhu, Rose,
and Han [22] 2018. Our results in [22] confirm both theoretically and experimentally the feasibility of
using FBG sensors for SH wave detection with a maximum reception efficiency around 45◦ incidence,
which shows a completely different angular dependence in comparison to Lamb-type wave detection.
In this current paper, taking advantage of the remote sensing (attenuation of the optical signal in
an optical fiber is ultralow) and multiplexing capability of FBG sensors, the work is expanded upon our
previous work in [22] to demonstrate how the technique could be used to extend the range in medium
and long-range pipe inspection applications using torsional guided waves. The use of the fundamental
shear horizontal mode in plates and torsional mode in pipes is of interest for ultrasonic inspection due
to the nondispersive nature and insensitivity to water loading that these modes offer. Since long-range
guided wave pipe inspection is critical and taking place all over the world on thousands of miles
of pipe, the work reported in this paper was intended to investigate if improvements in inspection
reliability and cost reduction would be possible with a hybrid ultrasonic guided wave-FBG approach.
Some earlier work on this subject was reported by Hu, Zhao, and Li [23] 2012 and Ray, Srinivasan,
Balasubramaniam, and Rajagopal [24] 2017. Hu et al. [23] 2012 performed a modeling study suggesting
feasibility to integrate optical fiber sensors into pipeline inspection using the L(0,2) mode, but an
experiment was not reported. Ray et al. [24] 2017 describe the use of fiber optic sensors to detect
changes in the guided wave features of the L(0,2) mode to assess cross sectional irregularities such as
pipe eccentricity. Obviously, their work does not address torsional wave detection. The work reported
16
Appl. Sci. 2018, 8, 727
in this paper will illustrate how FBG receivers could be implemented as part of a hybrid ultrasonic
pipe inspection technology using torsional guided waves.
2. Hybrid Ultrasonic—Fiber Optic System Concept

Commercially available systems for ultrasonic guided wave pipe inspection can quickly and
reliably scan large sections of pipeline for flaws. However, there are certain situations where increased
range of an inspection system is desired. This may occur in such situations as for highly attenuative
pipe, such as buried, coated, or generally corroded lines, or in situations where access is limited.
In these situations, use of an alternate method, such as fiber optic sensors, in conjunction with
conventional ultrasonic guided wave inspection can be beneficial to increase inspection range into
the non-inspected region. Fiber optic sensors may be attached to a pipeline during installation and
then left in place for use during periodic inspections. The fiber optic sensors could be placed near the
end of the range of a conventional pulse-echo inspection system, allowing for increased range by as
much as 50%, if the optical system has a similar signal-to-noise ratio. This can be achieved because the
reflected wave energy from a flaw need not travel the full return distance to the actuator, but only back
to the remote fiber optic sensor to be detected.
This concept is illustrated in Figure 1. The conventional ultrasonic collar generates guided waves
in the pipe at location 0. Remote fiber optic sensor(s) are placed at a distance X from the collar,
which should be near, but not beyond, the effective inspection range of the conventional system.
The inspection range of the hybrid system is then extended such that a flaw at a distance Y from the
collar is now detectable from the original collar location through use of the fiber optic sensor. Note that
the optical system could be collocated with the ultrasonic equipment, or not. One key difference
between optical fiber and conventional metal cable carrying ultrasonic signals is that the loss of the
optical signal is very low in comparison.
Figure 1. Global hybrid ultrasonic guided wave and fiber optic inspection system concept. The range
of a conventional collar at position 0 can be increased beyond position X using a fiber optic sensor
attached near position X. This allows the hybrid system to detect a flaw at a further distance, position Y.
Consider an example coated pipeline with high attenuation. It is possible that a conventional
inspection distance is only 50 feet. Flaws beyond 50 feet from the collar are not detected because
potential flaw reflections are hidden within the noise. However, if a fiber optic sensor was installed
near the 50-foot location, the inspection range could be increased by 25 feet, to a total of 75 feet.
Assuming constant attenuation along the pipe, the ultrasonic wave energy would decrease the same
amount traveling from 50 feet back to the collar, or from 50 to 75 feet and back to the fiber optic
sensor. If the fiber optic sensor has a similar signal-to-noise ratio as the ultrasonic collar, the effective
inspection range can be increased by as much as 50%. The fiber optic sensors could be attached and left
in place (under coatings) during pipeline installation, saving significant time and cost from inspection
and preparation costs to expose the pipe for collar application, particularly for the case where access
must be dug in the middle of a buried section of pipeline. Note that the use of a hybrid system could
be costly due to the need for extra components to perform optical sensing such as the laser, controller,
photodetector, etc. However, in certain cases where it can achieve improved performance, such as in
cases where increased inspection range provides inspection of previously un-inspectable or expensive
17
Appl. Sci. 2018, 8, 727
or hard-to-inspect areas, use of a hybrid system will be worthwhile. Future developments will also
lead to reduced cost of optical components. Generally, it is expected that the fiber sensors will be
used in a similar method as for conventional piezoelectric and magnetostrictive sensors, i.e., to detect
reflections from flaws. Although it may be possible to detect flaws between a sender and receiver,
flaw detection is much easier and more straight forward when analyzing a reflected wave signal from
a flaw.
Note also that additional SMS collar(s) could be installed at location X, etc., to extend the inspection
range, but these collars are generally more expensive than optical fiber, would require many, long
cables (typically 16, one per channel) with a remotely used data acquisition system, and there may not
physically be access space to install these collars. The use of fiber optic sensors also allows for many
sensors along a single optical cable.
3. Materials and Methods

An experiment was conducted on a 21-foot section of 6-inch NPS schedule 40 carbon steel pipe.
A segmented magnetostrictive (SMS) collar was placed 3 feet from the pipe end and used to generate
torsional axisymmetric ultrasonic guided waves in the pipe. Generation of ultrasound using the
magnetostrictive effect is described in Ref. [25] 1993. The SMS collar was constructed of copper
meandering coils printed on flexible circuit boards with a design frequency of 64 kHz, which were
placed over a 2-inch-wide strip of iron cobalt (FeCo) that was bonded to the circumference of the pipe.
Olympus UltraWave LRT hardware (OSSA, Waltham, MA, USA) and custom software were used
to generate torsional mode ultrasonic guided waves and collect data. This sending setup, and the
SMS coil design, is identical to the one used and described in Ref. [11]. Optical FBG sensors were
initially collocated with the collar, on the top and bottom of the pipe, and used to receive ultrasonic
guided wave energy. The optical fiber sensors were oriented at 45◦ relative to the pipe axis and directly
attached to the pipe surface using clear Scotch® tape. Tape was used because a temporary method
for sensor bonding was desired so that the fiber optic sensor could be easily and quickly moved to
various locations during experimental testing. Epoxy or superglue (cyanoacrylate) would be a more
appropriate long-term choice for permanent sensor installation. The 45◦ orientation was previously
shown to result in maximum reception of shear horizontal-type guided wave energy in plates [22]
2018. Ref. [22] also gives a theoretical justification for why this angle achieves the highest amplitude
response to shear-type guided waves. Two flaws were introduced in the pipe at 13 feet and 10 feet
from the pipe end. A schematic of the pipe is illustrated in Figure 2.
Figure 2. Schematic of the 21-foot pipe specimen. The SMS collar and the initial FBG sensor locations
were 3 feet from the pipe end. Flaws were introduced at 13 feet and 10 feet from the pipe end. The side
with flaw 2 will be referred to as the right side. Number and locations of the optical fiber sensors vary
in the different experiments.
A schematic of the test equipment is shown in Figure 3. A laptop was used for data acquisition
and custom software was used to control the UltraWave pulser/receiver hardware. The pulse was sent
to the collar to generate ultrasonic guided waves in the pipe. One or more fiber optic sensors were
coupled to the pipe and a distributed feedback (DFB) laser was used with a locking system to keep
the laser wavelength locked onto the spectral notch of the optical fiber sensor, and a photodetector
was used to convert the light signal into a voltage signal for recording. Different from the benchtop
18
Appl. Sci. 2018, 8, 727
narrow linewidth laser used in our previous work [22], the much more portable and low-cost DFB
lasers (A1905LMI, Avanex, New York, NY, USA) were used in this paper with moderate compromise
in signal-to-noise ratio (due to the larger noise of a DFB laser). Furthermore, a two-channel system was
used in this work to demonstrate the multiplexing capability of the FBG sensing system (see Figure 3),
i.e., one DFB laser and FBG pair works at 1550 nm wavelength while the other pair works at 1545 nm.
A dense wavelength division multiplexer (DWDM) was used to optically separate the two channels and
the ultrasonic signals received by the two FBG sensors are acquired simultaneously. Additional details
describing each single-channel setup are found in Ref. [22] 2018. Although the two-channel system
is enough for the demonstration of multiplexing capability, when using this type of optical setup
additional lasers could be added for additional simultaneous reception by fiber sensors connected to
other channels of the DWDM but it is not the focus of this paper. Note that the type of optical fiber
sensor used in this study is a Fabry-Perot interferometer formed by a pair of tandem chirped fiber
Bragg gratings, each with a 5-mm grating length, abbreviated as CFBG-FPI and referred to in this work
simply as an optical fiber sensor. Any reasonable optical fiber Bragg grating type sensor may be used
to detect torsional guided waves; this sensor type was selected due to its relatively high signal-to-noise
ratio, although it has other properties such as the ability to withstand relatively large static strains in
the structure while still using its narrow spectral notch pattern [26] 2017.
Figure 3. Schematic of the integrated ultrasonic-optical system setup. A servo controller was used to
keep the laser wavelength locked to the slope of the FBG spectrum for each channel. The optical signal
was converted to a voltage signal and recorded using conventional ultrasonic hardware.
4. Results
4.1. Confirmation of Torsional Mode and Flaw Detection

Figure 4 shows a frequency scan (F-Scan) plot of data collected using ultrasound generated by the
SMS collar and received by the optical fiber sensor attached on the bottom of the pipe at 3 feet from the
pipe end. The F-Scan plot shows data from a series of A-scans (amplitude scans) with different pulse
center frequencies vs. distance, with amplitude shown in color. The dead zone, pipe end reflections,
and flaw indications are indicated. These two flaws are located at 7 feet and 10 feet from the sensor
and on the top and right side of the pipe, respectively. The flaws are 0.5-inch diameter and 0.25-inch
deep drilled holes, which represent a simulated 2% CSA (pipe cross-sectional area) corrosion flaw.
19
Appl. Sci. 2018, 8, 727
Reflections from a reverse wave are also shown in Figure 4. Although the SMS collar was activated
with directional control, in which two axially-separated coils are used with a time delay to cancel the
backward-travelling wave, the cancellation is not perfect, and therefore extra reflections are shown at
3 feet and 13 feet, and to a lesser extent near 10.5 feet, from the collar location. These signals would
not be as prevalent in a longer length of pipeline, since a 100% reflector (cut pipe end) would not be
present. Also, these reverse wave reflections could be suppressed if two axially-separated optical fiber
sensors are used and a directional control routine is applied on the received signals.
Figure 4. F-Scan generated using an FBG sensor directly bonded to the bottom side of the pipe at 3 feet
from the pipe end, collocated with SMS sending collar. Flaw indications are shown at 7 feet and 10 feet
from the sensor. Pipe end reflections are shown at −3 feet and 18 feet. Indications at 3 feet and 13 feet
are due to waves traveling from the SMS collar in the reverse direction and bouncing off the near end
of the pipe at −3 feet, a result of imperfect directional control. The optical fiber sensor can clearly be
used for flaw detection.
Even though the FBG sensor is effectively sensing at only a small portion (approximately a
“point”) along the circumference, as opposed to sensing around nearly the full circumference as is
achieved with a conventional long-range ultrasonic collar, it is clear from Figure 4 that positive flaw
detection occurs over a range of frequencies. The ability of the “point” optical fiber sensor to detect
a flaw reflection is dependent on the relative locations of the flaw and sensor, as will be discussed
further below.
For comparison, F-Scan data using solely the SMS collar in pulse-echo mode is displayed in
Figure 5. The two 2% CSA flaws are clearly detectable at 7 and 10 feet. Also, there is a higher
signal-to-noise ratio for this setup, which is typically seen for magnetostrictive sensors. Use of different
optical components, such as a laser with lower noise, could achieve a similar signal-to-noise ratio.
Note also that the dual axially-separated coils allow for directional control and also a suppression
algorithm on the received wave, suppressing the reverse waves shown in Figure 4, which is not possible
for a single fiber optic sensor but could be achieved with multiple fiber optic sensors, as noted above.
The experimental pipe tested here does not highlight the advantages of the hybrid sensing method,
since this application of a short pipe of low attenuation would not benefit from range extension.
The overall goal of this work is to prove that detection with a hybrid system is possible and describe
future applications where it could prove beneficial.
20
Appl. Sci. 2018, 8, 727
Figure 5. F-Scan generated using the SMS collar in pulse-echo mode at 3 feet from the pipe end.
Flaw indications are shown at 7 feet and 10 feet from the sensor. Pipe end reflections are shown at
−3 feet and 18 feet. There is a higher signal-to-noise ratio for this setup. Note also that the dual coils
allow for directional control of the received wave, suppressing the reverse waves shown in Figure 4.
4.2. Progressive Damage Test in SHM Mode

Hybrid ultrasonic-optical data was also collected in a structural health monitoring (SHM) mode.
In SHM mode, data was collected on the undamaged pipe, and then after each successive growth
in flaw size. Figure 6 shows enveloped waveforms from several SHM states when the second flaw,
located 7 feet from the sensor, was at 0%, 0.5%, 1% and 2% CSA. As shown, the reflection amplitude
increases as the flaw size increases. Additionally, the reflection amplitude from the flaw at 10 feet from
the sensor remains relatively constant, because the size of this flaw remains the same. Note that the
0.5% and 1% CSA flaws represent a 0.25-inch diameter hole with 0.125-inch depth and a 0.25-inch
diameter hole with 0.236-inch depth, respectively.
Figure 6. Waveform data (amplitude envelope) for SHM states with flaw growth of 0 to 2% CSA for
the flaw located 7 feet from the optical fiber sensor. Waveforms are shown for an excitation frequency
of 76 kHz using the SMS collar and a single direct-bonded FBG optical fiber sensor used as a receiver.
As the flaw grows, the reflected ultrasonic wave energy from the flaw increases. The optical fiber sensor
can clearly be used for flaw sizing.
4.3. Individual FBG Sensor NDT

It is clear from Figures 4 and 6 that the ultrasonic-optical approach can be effective for flaw
detection. This section assesses the ability of a point-like receiver to reliably detect reflections from a
particular flaw location. What is found is that if the flaw is relatively far from the receiving sensor, then
detection is possible at all circumferential locations (see Figure 7). However, if the flaw is relatively
close to the receiving sensor, then detection is less reliable as it is more dependent on the circumferential
21
Appl. Sci. 2018, 8, 727
location of the FBG relative to the flaw (see Figure 8). This result will be dependent on the scattering
field from the flaw at small sensor-flaw distances, which will vary according to flaw geometry and
ultrasonic wavelength. However, at further distances, for circular and many corrosion-like flaws,
the reflected wave will be detectable at nearly any circumferential location due to the wrapping of the
reflected wave path circumferentially around the pipe. Additionally, sweeping inspection frequency to
generate the F-scan further improves the likelihood of detection.
Figure 7. F-Scans collected using a direct-bonded FBG receiver collocated with the sending collar
at several circumferential locations: (a) top; (b) right side; (c) bottom; and (d) left side of the pipe.
Since the receiver is relatively far from the flaws, both flaws (red ovals) are clearly detectable at all
circumferential locations. Flaw indications are marked by the red ovals. Arrows indicate the axial
location of the fiber optic receiver. Distance is indicated from source/receiver position. When the
receiver is relatively far from the flaw, i.e., multiple pipe circumferences, flaw detection is reliable at
any circumferential location. Note that the indication in (c) at 13 feet (rectangle) is due to a reverse
wave reflection.
22
Appl. Sci. 2018, 8, 727
Figure 8. F-Scans collected using a direct-bonded FBG receiver 6 feet in front of the sending collar
at several circumferential locations: (a) top; (b) right side; (c) bottom; and (d) left side of the pipe.
Since the receiver is relatively close to the flaws, the best flaw detection occurs directly in line with the
flaw location, while the detection at other circumferential locations is unreliable. Clear flaw indications
are marked by the red ovals and less clear or undetectable flaw indications are marked by dashed
red ovals. Arrows indicate the axial location of the fiber optic receiver. Distance is indicated from
the sending collar. When the receiver is near the flaw, but not at the same circumferential location,
flaw detection is not reliable, due to the use of a point-like sensor.
Figure 7 shows F-Scans collected with optical fiber sensors collocated with the collar at four
distinct circumferential positions (top, right, bottom, and left sides). The axial distance to the two flaws
is 7 feet and 10 feet, respectively. Compared to the pipe circumference, this axial distance is relatively
large compared to the pipe circumference, at 4 and 5.8 times the circumference, respectively. At this
axial distance, sensors at all four circumferential positions detect both flaw reflections. It is believed
that detection is possible at all circumferential locations because the scattered wave generated by the
23
Appl. Sci. 2018, 8, 727
interaction of the initial axisymmetric wave and the flaw has had sufficient time to spiral around the
pipe by the time the waves return to the receiving sensor locations.
Figure 8 shows F-Scans collected with optical fiber sensors located 6 feet in front of the collar at
four separate circumferential positions. In this setup, the axial distance to the two flaws is 1 foot and
4 feet, respectively. This is a relatively small axial distance between the sensors and flaws at 0.6 and
2.3 times the circumference, respectively. In this case, the sensor located on the opposite side of the pipe
from the flaw has the lowest reflected wave amplitude from that position. In both cases, the sensor on
the opposite side does not effectively detect the flaw, and the reliability of the other sensor locations for
flaw detection is intermittent. This may be due to the smaller amount of energy in the scattered wave
field travelling at high angles relative to the axis of the pipe, and also the natural angular reception
efficiency of the optical sensor which decreases as incident angle increases relative to the pipe axis,
with a maximum efficiency at 45◦ . The exception is for the sensor at the same circumferential location
(directly in line) with the flaw. The sensor on the top of the pipe clearly detects the flaw at 10 feet,
and the sensor on the right of the pipe clearly detects the flaw at 7 feet. These results suggest that to
have 100% detection coverage of the pipe using point-like sensors, including at axial distances close to
the sensors, an array of optical fiber sensors is preferred—either circumferentially around the pipe,
axially around the pipe, or a combination of the two. An array of sensors can detect flaws regardless of
the closeness of the optical sensors to the flaws.
4.4. Sparse Array NDT

As mentioned at the end of the previous section, to increase flaw detection reliability, a sparse
array of optical fiber sensors could be situated at different circumferential and/or axial locations
along the pipe. This would allow detection of a flaw at any location along a pipe, because if it is
close to one sensor it will also be sufficiently far from other sensors to allow for reliable detection.
Several sparse arrays were simulated by moving the optical fiber sensors to different locations on the
test pipe. An optical fiber sensor was located at 0 feet, 3 feet, 6 feet and 9 feet from the sending collar
on the top, left, bottom, and right sides at each of those axial locations. Data was collected at two
locations at a time until data from all 16 locations was collected. Data from multiple positions is then
combined to allow for an assessment of a variety of sensor array geometries.
In the present work, two optical fiber sensor arrays are considered. Both spiral down the pipe so
that there is one sensor located on the top, right, bottom, and left sides of the pipe, each at a different
axial location. Group 1 contained sensors on the left, bottom, right, and top, and group 2 contained
sensors on the right, top, left, and bottom of the pipe at axial locations of 0 feet, 3 feet, 6 feet and 9 feet,
respectively. This is shown schematically in Figure 9. These groups were selected such that group 1
had sensors directly in front of the two flaws, while the sensors closest to the flaws in group 2 are on
the opposite side of the pipe (and therefore those two sensors are not expected to easily detect the
flaws). Note that the flaws are located on the top and right sides of the pipe at 10 feet and 7 feet from
the sending collar, respectively.
To combine the four waveforms collected via the four FBGs in each array into a single “group
waveform”, the individual waveforms are processed by suppressing known position-dependent signals
including the dead zone, the direct wave arrival, and the reverse pipe end reflection. When these
events are present in the signals, the amplitude is reduced to a value of 20 mV, chosen because it is
close to the noise level. The amplitude envelope of each of the four waveforms is also shifted in time
to correspond to their respective positions along the pipe axis. After time-shifting, the waveforms
are added, and the average is calculated. Flaws are more easily detectable in the group waveform
than in any individual waveform from a single FBG. Additionally, the combined waveform is more
robust than data from a single location since we are unlikely to miss a flaw due to a particular flaw
position relative to a single sensor. Note that in general, the arrival times of these suppressed features
would be known a priori since the sensor locations are known. This suppression algorithm also adds a
jaggedness to the waveforms, as the suppression is applied as a step function. For the data shown in
24
Appl. Sci. 2018, 8, 727
Figure 10, this jaggedness occurs for the approximate length of the initial wave pulse (1 foot) as the
wave passes each sensor location: 1–2 ft, 3–4 ft, 6–7 ft and 9–10 ft.
Figure 9. Schematic showing sensor locations for array groups 1 (above) and 2 (below). Fiber optic
sensors are indicated by green rectangles and the optical fiber is indicated by the blue line. Dashed lines
indicate the optical fiber and sensor on the far side of the pipe. The pipe is being viewed from the right
side (flaw 2 side).
Figure 10. F-Scans of the processed group waveforms for both (a) group 1 and (b) group 2 fiber optic
sensor arrays. Flaws (red ovals) are clearly detected at 7 feet and 10 feet from the sending collar across
a wide range of frequencies. The far pipe end reflection is detected at 18 feet. Note that groups 1 and 2
both detect the flaws even though group 2 has multiple sensors near flaws in non-ideal locations for
efficient flaw detection.
Figure 10 shows F-Scans of the processed group waveforms for both (a) group 1 and (b) group 2
fiber optic sensor arrays over 30 to 100 kHz frequencies generated with the SMS collar. The combined
25
Appl. Sci. 2018, 8, 727
group 1 and group 2 signals both show good flaw detection capabilities. Both FBG sensor groups
detect the flaws at 7 feet and 10 feet from the collar location, even though multiple sensors in group 2
are on the far side of the pipe from the flaw to which they are nearest. This confirms that the potential
to miss a flaw based on its relative location to a single sensor is greatly diminished using this approach
versus placing a single fiber sensor at a single axial location. The array approach can increase detection
confidence. Creating a group signal from sensors at 4 circumferential locations at a single axial location
can also give a similar result. Depending on inspection requirements and access limitations, different
implementations of a sensor array may be desired. Additionally, the array group signals appear to
be more stable over a wider frequency range, which makes for more reliable interpretation by the
inspector. Due to these characteristics, the F-Scans in Figure 10 are easier to interpret and more reliable
for flaw detection than the F-Scans shown in Figures 7 and 8, for a single fiber sensor at a single axial
and circumferential location.
5. Discussion
It is believed that the work presented here represents the first study of flaw detection in pipes using
an ultrasonic-optical system and torsional ultrasonic guided waves. Torsional wave reception using
FBG sensors was successful, and damage detection, monitoring, and localization were demonstrated.
Additionally, one method for increased confidence for flaw detection was demonstrated, in which
signals from a sparse array of optical fiber sensors were combined.
A hybrid ultrasonic-optical inspection system has a wide variety of potential applications.
One application is with permanently-mounted long-range ultrasonic guided wave pipe inspection
collars. The optical fiber could be far from the excitation area and can increase the effective inspection
range from a single sending collar position. Optical fiber sensors may also be used in places with
limited access for traditional collars. The optical cable is very thin (as little as 80 μm diameter for
bare fiber), so it can easily be mounted under coatings. Additionally, the optical fiber sensors can
withstand harsh environments such as high temperature, radiation, and high levels of electromagnetic
interference (EMI). Finally, the optical fiber sensors have a much wider bandwidth than traditional
piezoelectric or magnetostrictive sensors.
Future work should focus on the development of ruggedized and miniaturized optical
components and the demonstration of system effectiveness in field tests. Tests demonstrating
an increase in inspection range on buried pipe could open avenues for additional applications.
Optical fiber sensors provide unique advantages over traditional sensor types which will encourage
their use in challenging inspection problems in the future.
Author Contributions: Joseph L. Rose and Jason Philtron contributed the ultrasonic guided wave and hybrid
aspect of the research work including theory, experiment, and the write up. Guigen Liu, Yupeng Zhu,
and Ming Han contributed fiber optic Bragg grating experimental portion of the work and assisted in
manuscript revision.
Funding: The work carried out by J. L. Rose and J. Philtron was funded by Naval Research Laboratory grant
number N0001417RX00182. The work carried out by G. Liu, Y. Zhu, and M. Han was carried out primarily at
University of Nebraska-Lincoln and funded by the Office of Naval Research under Grant N000141712819 and
Grant N000141410456.
Acknowledgments: The authors would like to thank Bob Brown and James Tagert at the Naval Research
Laboratory for making this work possible. The authors would also like to acknowledge Russell Love and
Cody Borigo for assistance with software and the experimental setup.
References
1. Rose, J.L. Ultrasonic Waves in Solid Media; Cambridge University Press: New York, NY, USA, 1999.
2. Rose, J.L. Ultrasonic Guided Waves in Solid Media; Cambridge University Press: New York, NY, USA, 2014.
26
Appl. Sci. 2018, 8, 727
3. Ditri, J.J.; Rose, J.L.; Chen, G. Mode Selection Guidelines for Defect Detection Optimization Using Lamb
Waves. In Proceedings of the 18th Annual Review of Progress in Quantitative NDE, Brunswick, ME, USA,
28 July–2 August 1991; Plenum: New York, NY, USA, 1991; Volume 11, pp. 2109–2115.
4. Pilarski, A.; Ditri, J.J.; Rose, J.L. Remarks on Symmetric Lamb Waves with Dominant Longitudinal
Displacements. J. Acoust. Soc. Am. 1993, 93, 2228–2230. [CrossRef]
5. Rose, J.L.; Soley, L.E. Ultrasonic Guided Waves for Anomaly Detection in Aircraft Components. Mater. Eval.
2000, 50, 1080–1086.
6. Rose, J.L.; Sun, Z.; Mudge, P.J.; Avioli, M.J. Guided Wave Flexural Mode Tuning and Focusing for Pipe
Inspection. Mater. Eval. 2003, 61, 162–167.
7. Hayashi, T.; Song, W.J.; Rose, J.L. Guided Wave Dispersion Curves for a Bar with an Arbitrary Cross-section,
a Rod and Rail Example. Ultrasonics 2003, 41, 175–183. [CrossRef]
8. Gao, H.; Shi, Y.; Rose, J.L. Guided Wave Tomography on an Aircraft Wing with Leave in Place Sensors.
In Review of Quantitative Nondestructive Evaluation; AIP: New York, NY, USA, 2005; Volume 24, pp. 1788–1794.
9. Lee, C.M.; Rose, J.L.; Cho, Y. A Guided Wave Approach to Defect Detection under Shelling in Rail. NDT E Int.
2008, 42, 174–180. [CrossRef]
10. Puthillath, P.; Rose, J.L. Ultrasonic Guided Wave Inspection of a Titanium Repair Patch Bonded to an
Aluminum Aircraft Skin. Int. J. Adhes. Adhes. 2010, 30, 566–573. [CrossRef]
11. Van Velsor, J.; Royer, R.; Owens, S.; Rose, J.L. A Magnetostrictive Phased Array System for Guided Wave
Testing and Structural Health Monitoring of Pipe. Mater. Eval. 2013, 71, 1296–1301.
12. Philtron, J.H.; Rose, J.L. Mode Perturbation Method for Optimal Guided Wave Mode and Frequency Selection.
Ultrasonics 2014, 54, 1817–1824. [CrossRef] [PubMed]
13. Rose, J.L.; Borigo, C.; Owens, S.; Reese, A. Rapid Large Area Inspection from a Single Sensor Position:
A Guided Wave Phased Array Scan. Mater. Eval. 2017, 75, 671–678.
14. Chen, L.; Choi, W.R.; Lee, J.G.; Kim, Y.G.; Moon, H.S.; Bae, Y.C. Oil-Tank Weld Detection Using EMAT. Int. J.
Hum. Robot. 2017, 14, 1750008. [CrossRef]
15. Kim, Y.G.; Chen, L.; Moon, H.S. Numerical Simulation and Experimental Investigation of Propagation of
Guided Waves on Pipe with Discontinuities in Different Axial Angles. Mater. Eval. 2016, 74, 1168–1175.
16. Chen, L.; Kim, Y.G.; Bae, Y.C. Long Range Displacement Measurements Systems Using Guided Wave. Int. J.
Fuzzy Log. Intell. Syst. 2017, 17, 154–161. [CrossRef]
17. Perez, I.; Cui, H.-L.; Udd, E. Acoustic emission detection using fiber Bragg gratings. In Proceedings of the
SPIE’s 8th Annual International Symposium on Smart Structures and Materials, Newport Beach, CA, USA,
4–8 March 2001; International Society for Optics and Photonics: Bellingham, WA, USA, 2001.
18. Wild, G.; Hinckley, S. Acousto-ultrasonic optical fiber sensors: Overview and state-of-the-art. IEEE Sens. J.
2008, 8, 1184–1193. [CrossRef]
19. Li, F.; Murayama, H.; Kageyama, K.; Shirai, T. Guided wave and damage detection in composite laminates
using different fiber optic sensors. Sensors 2009, 9, 4005–4021. [CrossRef] [PubMed]
20. Harish, A.V.; Ray, P.; Rajagopal, P.; Balasubramaniam, K.; Srinivasan, B. Detection of fundamental shear
horizontal mode in plates using fibre Bragg gratings. J. Intel. Mater. Syst. Struct. 2016, 27, 2229–2236.
[CrossRef]
21. Ray, P.; Rajagopal, P.; Srinivasan, B.; Balasubramaniam, K. Feature-guided wave-based health monitoring of
bent plates using fiber Bragg gratings. J. Intel. Mater. Syst. Struct. 2017, 28, 1211–1220. [CrossRef]
22. Liu, G.; Philtron, J.H.; Zhu, Y.; Rose, J.L.; Han, M. Detection of Fundamental Shear Horizontal Guided Waves
Using a Surface-bonded Chirped-fiber-Bragg-grating Fabry-Perot Interferometer. IEEE J. Lightwave Technol.
2018, 36, 2286–2294. [CrossRef]
23. Hu, Y.; Zhao, M.; Li, S. Pipeline Defect Detection Based on Ultrasonic Guided Wave Technique Using Fiber
Bragg Gratings. In Proceedings of the International Conference on Pipelines and Trenchless Technology
(ICPTT), Wuhan, China, 19–22 October 2012; pp. 996–1010.
24. Ray, P.; Srinivasan, K.; Rajagopal, P. Fiber Bragg Grating-based Detection of Cross Sectional Irregularities
in Metallic Pipes. In Proceedings of the SPIE, 25th International Conference on Optical Fiber Sensors, Jeju,
South Korea, 24–28 April 2017; pp. 1–4.
27
Appl. Sci. 2018, 8, 727
25. Trémolet de Lacheisserie, E. Magnetostriction: Theory and Applications of Magnetoelasticity; CRC Press:
Boca Raton, FL, USA, 1993; Volumes 339–352, pp. 359–361.
26. Zhang, Q.; Zhu, Y.; Luo, X.; Liu, G.; Han, M. Acoustic Emission Sensor System Using a Chirped
Fiber-Bragg-grating Fabry-Perot Interferometer and Smart Feedback Control. Opt. Lett. 2017, 42, 631–634.
[CrossRef] [PubMed]
28
applied
sciences
Article
Enhancement of Ultrasonic Guided Wave Signals
Using a Split-Spectrum Processing Method
Seyed Kamran Pedram 1, *, Peter Mudge 1 and Tat-Hean Gan 1,2
1 TWI Ltd., Granta Park, Great Abington, Cambridge CB21 6AL, UK; peter.mudge@twi.co.uk (P.M.);
Tat-Hean.Gan@brunel.ac.uk (T.-H.G.)
2 Brunel University London, Institute of Materials and Manufacturing, Kingston Lane, Uxbridge UB8 3PH, UK
* Correspondence: kamran.pedram@twi.co.uk
Received: 10 August 2018; Accepted: 29 September 2018; Published: 3 October 2018
Abstract: Ultrasonic guided wave (UGW) systems are broadly utilised in several industry sectors
where the structural integrity is of concern, in particular, for pipeline inspection. In most cases,
the received signal is very noisy due to the presence of unwanted wave modes, which are mainly
dispersive. Hence, signal interpretation in this environment is often a challenging task, as it degrades
the spatial resolution and gives a poor signal-to-noise ratio (SNR). The multi-modal and dispersive
nature of such signals hampers the ability to detect defects in a given structure. Therefore, identifying
a small defect within the noise level is a challenging task. In this work, an advanced signal
processing technique called split-spectrum processing (SSP) is used firstly to address this issue
by reducing/removing the effect of dispersive wave modes, and secondly to find the limitation
of this technique. The method compared analytically and experimentally with the conventional
approaches, and showed that the proposed method substantially improves SNR by an average of
30 dB. The limitations of SSP in terms of sensitivity to small defects and distances are also investigated,
and a threshold has been defined which was comparable for both synthesised and experimental data.
The conclusions reached in this work paves the way to enhance the reliability of UGW inspection.
Keywords: signal processing; SNR; split-spectrum processing; ultrasonic guided waves
1. Introduction
Long-range ultrasonic testing (LRUT), also known as guided wave testing (GWT) is an advanced
non-destructive testing (NDT) method that utilises ultrasonic guided wave (UGW) signals for the
inspection. This inspection could be applied to any large complex structures such as pipes, rails,
cables, etc. for defect detection. This method is widely utilised for the inspection of pipelines that
mainly contain oil and gas products, and it has the ability to screen long distances (up to 50 m in each
direction) from a single location to identify defects in the structure (e.g., corrosion, erosion) [1–3]. GWT
often operates at a low-frequency range (20–100 kHz) (compared to conventional ultrasonic testing
(UT), which operates at MHz range) to transmit the waves using one or more rings of dry-coupled
transducers around the circumference of the pipe, which are pneumatically forced against the surface.
These waves propagate within the pipe wall along the pipe’s main axis, and scattering occurs when the
waves encounter discontinuities in wall thickness. The transducers are used to record these changes to
obtain information about the presence and characteristics of the features within a pipe [4].
In order to reduce the effect of dispersion and to achieve a good resolution between features,
a tone burst signal is employed for the transmission of the signal, as shown in Figure 1a.

Appl. Sci. 2018, 8, 1815
D
Figure 1. guided wave testing (GWT) signals: an excitation (a) time domain and (b) frequency domain
signal, received (c) time domain and (d) frequency domain signal.
This is a 50 kHz 5-cycle Hann windowed excitation signal, with a frequency response as shown in
Figure 1b, which are generated in this work using MatLab software. A typical response is illustrated in
Figure 1c, which consists of a number of peaks that correspond to reflections from structural features
under investigation (e.g., defects and welds). In addition, its frequency response is displayed in
Figure 1d, exhibiting the same frequency bandwidth as the input signal [5].
It is ideal to generate an axisymmetric wave mode to promote non-dispersive propagation;
however, the interaction of the guided wave signal with non-axisymmetric features within the pipeline
can cause mode conversion. This results in the generation of dispersive wave modes (DWM) that
travel with different velocities according to the different frequency components in the signal [6]. Hence,
the energy spreads over space during propagation, and compromises the ability to distinguish echoes
from closely spaced reflectors. This lack of spatial resolution leads to coherent noise and reduces the
sensitivity of the inspection. In order to enhance the sensitivity and the SNR of such signals, it is vital
to minimise the presence of coherent noise. Dispersion is one of the main sources of coherent noise;
hence, the aim of this work is to reduce the effect of dispersive wave modes.
There are many researchers who study the effect of dispersion in GWT [7–14]. Wilcox [7]
developed a method for reversing the effect of dispersion by using the knowledge of wave mode
characteristics to map signals from the time to the distance domain, and then reversed the effect
of DWM and restored them to undispersed pulses. Zeng and Lin [8] investigated the dispersion
pre-compensation method using chirp-based narrowband excitation signals to compress the time
duration of received wave packets during the extracting process. They employed the benefits of chirp
excitation by utilising previous knowledge of the dispersion curve and the propagation distance. Other
researchers such as Xu et al. [9–11] employed dispersion compensation (DC) to analyse the propagation
behaviours of the signals. However, most of them required the knowledge of the propagation distance
in advance. Toiyama and Hayashi [12] combined the DC method with a pulse compression (PuC)
algorithm by employing a chirp signal. They considered a scenario of a single wave mode without
introducing the quantitative SNR enhancement. The combination of DC with PuC is utilised by
Yucel et al. [13,14] to enhance the SNR of UGW response employing a broadband maximal length
sequence excitation signal. The result showed that the technique was successful for highly dispersive
flexural wave modes, but it was not that effective for longitudinal wave modes that are not dispersive.
Mallet et al. [5] considered cross-correlation and wavelet de-noising algorithms for reduction of the
effect of dispersion in UGW. He claimed that neither of these methods was suitable for the reduction
30
Appl. Sci. 2018, 8, 1815
of coherent noise, as both methods removed the smaller amplitudes regardless of whether or not they
were signal or noise.
Newhouse et al. [15] considered split-spectrum processing (SSP) in the field of NDT to enhance
the SNR by splitting the signal’s response into a set of sub-band signals. A theoretical basis for the
selection of filter bank parameters was investigated by Karpur et al. [16]. They proposed an equation
to predict SNR enhancement by compounding a number of frequency diverse signals. The result
showed that some parameters achieved a larger value than expected, which could be the result of
using the Gaussian function for filtering (because of its simplicity) while the calculation was based
on the Sinc function. Shankar et al. [17] employed a polarity thresholding (PT) algorithm for the
detection of a single target. They showed how sensitive the SNR enhancement was to the selection of
filter bank parameters. Saniie et al. [18] investigated the performance of order statistic (OS) filters in
conjunction with SSP in the context of ultrasonic flaw detection, to improve the flaw-to-clutter ratio of
backscattered signals. It has been claimed that the OS filter performs well where the flaw and clutter
echoes have good statistical separation in a given quartile. However, its performance deteriorated
with the contamination from unwanted statistical information.
Gustafsson and Stepinski [19] adapted an SSP method using an artificial neural network (ANN)
to implement PT for UT signals. In order to allow the relative importance of the different sub-bands
to be taken into account, weighting factors were added to the input signal. The results showed
better performance than, PT but only for one particular sample. Gustafsson [20] then extended the
method by employing both the filter bank and the non-linear processing as an ANN for SSP. This
method was time-consuming, although the results indicated that the ANN could “eliminate most
of the noise”. Rubbers and Pritchard [21] developed a complex-plane SSP (CSSP) method, which
was a modified version of SSP for the ultrasonic inspection of castings, and improved the SNR for a
number of conventional UT techniques. It utilised an additional mathematical dimension to improve
the result, while maintaining linearity in both the amplitude and the energy content of a defect signal.
Moreover, they gave an overview of SSP methods [22] with a variety of SSP reconstruction algorithms
and parameters. However, they claimed that as the amplitude of the processed signal is non-linear, it
does not allow for the sizing of flaws, hence the use of this method is limited.
Rodriguez et al. [23,24] proposed a new filter bank design for SSP, based on the use of variable
bandwidth filters, where filters were equally spaced in frequency and their energy gain equalised.
They utilised stationary models for the grain noise in the presence of a single defect. They claimed
that a frequency multiplication (FM) algorithm gave the greatest SNR enhancement. They stated that
the number of filter bands compared to other algorithms is reduced; hence, it reduced the system
complexity. However, this technique was not evaluated for non-stationary models, highly dispersive
material, or a model with multiple defects. Syam and Sadanandan [25] employed a combination of
SSP and order statistic filters to reduce the effect of reverberation for flaw detection in conventional UT
using a wideband signal. He stated that by processing the multiple echoes corresponding to a set of
transmitted signals, the effect of microstructure reflections could be suppressed with respect to the
flaw echo. However, this method tested for a simulated signal only, without revealing the value of the
SSP parameters.
Overall, it is clearly shown by some of the cited literature [7–14] that their techniques required
prior knowledge of the dispersion curve and/or propagation distance in order to perform well. Other
cited literature [15–25] indicated that the successful implementation of the SSP technique was highly
sensitive to the selection of filter bank parameters. Although most of these papers claimed that they
enhanced the SNR of the signal’s response, the enhancement was mainly achieved for conventional
UT. However, those parametric values are not suitable for use in GWT that contain a combination
of axisymmetric and non-axisymmetric wave modes with different phase velocities. Hence, a full
study is required to find the optimum filter bank parameters for SSP in terms of its capacity to provide
such improvements in GWT. To the best of the authors’ knowledge, prior to this study, apart from
Mallet et al. [5], no one else has investigated the use of SSP in guided wave testing.
31
Appl. Sci. 2018, 8, 1815
An analysis of SSP with application to GWT was conducted in our previous paper [26] for
reducing the effects of DWMs in the signal response, and the optimum parameters have been proposed
to enhance the SNR and spatial resolution of such signals. However, the limitations of SSP for use
in GWT are still unclear, and they need further investigation. Hence, in this paper, the optimum
parameters that were identified in the previous work are utilised for deeper investigation to improve
the SNR, and to enhance the spatial resolution further by investigating the limitations of SSP mainly in
two areas: (i) in terms of defect sensitivity and (ii) in terms of minimum distance between two features
(e.g., weld and defect). Thus, the core concept explored in this work is to address these issues. Such a
parametric study has not been undertaken in the field of GWT prior to this work.
In order to do this, a synthesised signal has been created to identify the limitations of SSP.
The limitations have been tested, evaluated, and identified synthetically in a controlled environment.
Then, in order to validate the effectiveness of the technique, laboratory experiments have been carried
out on an eight-inch pipe utilising the Teletest-guided wave system [27]. It is shown that the proposed
method reduces the presence of coherent noise and improves the SNR by up to 30 dB. In addition,
the limitations of SSP have been identified and a threshold has been defined, below which the temporal
resolution will be reduced.
The paper is organised as follows: Section 2 describes the theory and concept of SSP, including
the implementation and selection of filter bank parameters for guided wave testing. Sections 3 and 4
provide details and discussion of SSP for synthesised and experimental testing, and finally Section 5
concludes the paper.
2. Split-Spectrum Processing (SSP)
2.1. Theory of SSP

SSP is an advanced signal processing technique that was initially developed from the frequency
agility techniques used in radar [28]. This method was then considered for SNR enhancement in NDT
applications such as conventional UT, to reduce grain scatter in the received signal. A significant
amount of research has been undertaken over the last few decades in this area with respect to the
reduction of non-random noise (coherent noise) in NDT applications, due to ultrasonic scattering.
The application of SSP in GWT is relatively new and, to the best of the authors’ knowledge, this
technique and its limitations have not been previously investigated in this field.
SSP splits the spectrum of a received signal in the frequency domain using a bank of bandpass
filters to generate a set of sub-band signals at incremental centre frequencies. These sub-band signals
are normalised and then subjected to a number of possible non-linear processing algorithms to generate
an output signal. Figure 2 illustrates a block diagram for SSP with its step-by-step implementation. It
shows that the input time domain signal, x(t), is transformed into the frequency domain, X(f), using
a fast Fourier Transform (FFT), and is then filtered by a bank of band-pass filters. Subsequently, the
outputs from the filter banks, Xk (f) (k = 1, 2, . . . , n), are converted back into the time domain using
the inverse FFT (IFFT) and normalised by a weighting factor, wk . Then one of the recombination
algorithms will be employed to combine these non-linear signals to produce the output signal y(t). In
general, SSP application shows a great potential to reduce those signal components that vary across a
frequency range, in particular dispersive wave modes, and to suppress the regions of the signal of
interest that are constant in that frequency range [29].
32
Appl. Sci. 2018, 8, 1815
Figure 2. Split-spectrum processing (SSP) block diagram.
2.2. SSP Filter Bank Parameters

The SSP filter-bank parameters were first investigated by trial-and-error for NDT applications as
they were processed. However, this was not very practical for field inspection as there are typically
large amounts of data to analyse; therefore, researchers sought to find the optimum values for NDT
methods. Hence, the optimum values have been proposed, developed and examined for UT techniques.
However, these values are not suitable for GWT due to the long duration and narrow bandwidth of the
signal that operates in the kHz range, whereas traditional UT operates in the MHz range. Therefore,
further investigation was required to find an optimum value suitable for GWT. As a result, the rules
and key factors of parameter selection for conventional UT have been reviewed and then modified for
use in GWT. Figure 3 illustrates the scheme of SSP filtering. The parameters that need to be quantified
are listed below with a brief description:
1. The total bandwidth for processing (B); this needs to be large enough such that the reflections of
the signal from features in the specimen are constant across this range, and the reflections from
coherent noise vary. If the bandwidth is too large, then it may cause the features to be lost, as
at least one of the filter outputs will not contain the feature signal. Hence it reduces the spatial
resolution in the processing. In general, narrowband waveforms were used as the excitation
signals to reduce the effect of unwanted wave modes, and to suppress the dispersion effect in the
GWT response. Hence, the bandwidth of the transmitted signal could be employed as the total
bandwidth of processing.
2. The filter separation (F); is the distance between the sub-band filters. Karpur et al. [16] claimed
that the optimum spectral splitting could be attained by using the frequency-sampling theorem,
whereby the spectrum of a time-limited signal can be reconstructed from sample points in the
frequency domain separated by 1/T Hz, where T is the total duration of the signal. Note that
the Gaussian filter is employed for calculating the filter bank in practice, due to its simplicity,
whereas the Sinc function was utilised for actual calculation. Thus, the filter separation could be
calculated as F = 1/T.
3. The sub-band filter bandwidth (Bfilt ) is the width to be used for each filter in the filter bank. It was
recommended [15–17] that the value of the sub-band filter bandwidth needs to be set at three to
four times the filter separation. It should be noted that a bandpass filter could reduce the temporal
resolution of the signal. This is because reducing the bandwidth of a time-limited signal will
increase its duration. This means that the SSP filter bank needs to be selected precisely, otherwise
it could lead to a reduction in temporal resolution, as the pulses that correspond to reflections
from features spread out in time and mask one another. Moreover, the correlation between
33
Appl. Sci. 2018, 8, 1815
adjacent sub-bands could be affected by the overlap of the filters. This means the correlation
increases with an increase in overlaps. On the other hand, little or no overlap could lead to loss
of information. It is notable that the noise in adjacent filters needs to be uncorrelated and the
features should be correlated. Hence, the overlap needs to be selected somehow to minimise the
correlation between coherent noise regions in adjacent sub-bands without losing information.
4. The filter crossover point (δ); or a cutoff frequency at the edges, is a boundary in the frequency
response at which the energy flowing through the structure starts to decrease rather than
passing through.
5. The number of filters (N); is the number of sub-band filters that is required to be selected in order
to enhance the SNR and spatial resolution by correlating the signal of interest and minimising
the correlation between the coherent noise region in the adjacent sub-bands signal. Overall, these
parameters are dependent on each other, which means that their values have a direct effect on
other parameter values. Therefore, it is necessary to search for the optimum parameters and to
select them appropriately. As an example, increasing the number of filters (N) would be required
to increase the total bandwidth (B), or to reduce the filter separation (F), or a combination of both.
Thus, as is shown in Figure 3 the number of filters (N) could be calculated as below:
B
N= +1 (1)
F
Figure 3. Filter bank parameters of SSP.
2.3. Recombination Algorithms of SSP

There are numerous SSP recombination algorithms that could be employed for use in GWT to
reduce the effect of coherent noise which have been described in the literature [16,22]. In this paper,
the two most common ones that obtained the highest SNR and spatial resolution in the previous work
are explained in more detail:
The polarity thresholding (PT) that is expressed as:
⎧
⎪
⎨ x[m] if all xi [m] > 0, i = 1, . . . , n
yPT [m] = x[m] if all xi [m] < 0, i = 1, . . . , n (2)
⎪
⎩ 0 Othrewise
where yPT is the output result of the signal that is obtained after processing at m, n is the number of
filter band signals, x[m] is the unprocessed signal, and xi [m] care the sub-band signals. This method
34
Appl. Sci. 2018, 8, 1815
looks at the signal sub-bands at each sample time, and if the samples are all negative or all positive,
then the output is the unchanged input signal. Otherwise, the output is zero. This has the effect of
only passing the time samples where the polarity is not affected by the frequency. Thus, those sections
of the signal that are highly frequency-dependent must be removed. However, the amplitude of the
signal of interest needs to be larger than the amplitude of the coherent noise response, whereas if the
noise signal has greater amplitude, then it will change the signal’s sign.
Polarity thresholding with minimisation (PTM), which is defined as:
⎧
⎪
⎨ min(x[m]) if all xi [m] > 0, i = 1, . . . , n
yPTM [m] = min(x[m]) if all xi [m] < 0, i = 1, . . . , n (3)
⎪
⎩ 0 Othrewise
PTM is the combination of the minimisation and the PT algorithm where the output is the
minimum amplitude of PT when there is no change in polarity. This method takes the minimum
points that the PT algorithm has passed in order to suppress the noise even further than is obtained by
the PT algorithm. Since the variance of the points containing noise is generally larger than those that
are containing the signal of interest, the use of the minimisation algorithm reduces those that are the
result of noise. However, this method is less effective when the noise level is larger than the actual
signal’s amplitude; thus, reducing the noise level will significantly reduce the signal amplitude in
certain sub-bands, and gives minimum values for the PTM’s output in that region.
2.4. Implementation of the Filter Bank

A MatLab program has been written that takes an unprocessed signal in the time domain and
converts it to the frequency domain. It then filters the signal using a Gaussian bandpass filter to
generate a set of sub-band signals, and applies the recombination algorithms into these sub-bands.
The input is the signal to be filtered, with the upper and lower 3 dB cut-off frequencies. Therefore, the
lower cut-off frequency fl , and the higher cut-off frequency fh , for each sub-band filter, are calculated as:

fmin − B4filt n = 1
fln = (4)
fln−1 + F n = 2, 3, . . . N
fhn = fln + Bfilt n = 1, 2, . . . N (5)
where F is the filter separation, N is the number of filters, fmin is the lower cut-off frequency of B,
and Bfilt are the sub-band filters. The lower cut-off frequency for the first sub-band fl1 , needs to cover
the start point of the signal. The selections of these values are inspired by the values that have been
employed in UT, and then adjusted for the use in GWT using the brute force search algorithm. Table 1
shows the optimum values of SSP that were proposed in the previous paper [26]. These values are
employed in this paper to find the limitations of SSP. The performance of the proposed technique is
quantified by measuring the SNR, the spatial resolution, and the defect sensitivity of the output signal.
Table 1. Recommended values for SSP parameters.
SSP Parameters Symbols Recommended Values

Total bandwidth B 99% of total energy
Sub-band filter bandwidth Bfilt B/11
Filter crossover point δ Bfilt /3.5
Filter separation F 1 dB
Number of filters N B/F + 1
35
Appl. Sci. 2018, 8, 1815
3. Signal Analysis
This section shows how this methodology has been evaluated. Initially, signals have been
simulated using the technique that was presented by Wilcox [7] to represent the guided wave reflections
in a pipe with known dispersion characteristics. This allows the SSP limitation to be found under
controlled conditions. Secondly, experimental data was collected from an eight inch pipe that was
equivalent to the simulated pipe. This gives two benefits: (i) it allows the approach to be tested
under more realistic conditions; and (ii) the reusable SSP parameters across equivalent pipes could
be evaluated.
Signal Synthesis
The signal synthesis was utilised to generate the propagation of the dispersive wave modes in
time/space, based on applying a frequency-dependent phase shift to the wave packet of interest via
DISPERSE software (developed by Imperial College, London, UK) [30]. The core concept of this section
was to identify the limitation of SSP in terms of finding the smallest defect size that could be detected,
and to find the distance limitation when the location of the defect is close to a dominant feature with a
high amplitude, such as a weld. In order to achieve that, the axisymmetric T(0,1) wave mode is excited
with a 10-cycle pulse and a centre frequency of 44 kHz, using the synthesised model. It is assumed that
there are only two reflection echoes (i) from the defect, and (ii) from the pipe’s end. Since the proposed
method has already proved its capability to remove the DWM [26], here, it is assumed that only the
T(0,1) wave mode is reflected from the features on the pipe. According to Böttger et al. [31], there is a
linear relationship between the amplitude of the reflected signal of T(0,1) and the cross-section area
(CSA) of its defect. Hence, the attenuation of T(0,1) is linear, which means that if 10% of the excited
signal reflects from the defect, then the rest of the energy (90%) will reflect from the pipe end.
The set-up of this experiment is shown in Figure 4. The distances of the defect and pipe end are
X = 3 m and X = 4.5 m from the excitation point, respectively. It is assumed that the defect reflects 6%
of the total energy and rest of the energy (94%) is reflected by the pipe end, as displayed in Figure 5a.
The reflections from the defect were reduced gradually in order to find the smallest defect size that can
be recognised by the proposed method. These are illustrated in Figure 5, where the defect sizes are
gradually reduced from 6% CSA to 1% CSA. The results demonstrate that the proposed method has
the potential to detect defects down to 1% CSA when the distance between two features is 1.5 m. Then,
in order to find the distance limitation, it is assumed that a 10% CSA defect is 1 m from the pipe end
and it is moved towards the pipe end in steps of 0.1 m as illustrated in Figure 4. Results in Figure 6a–d
clearly demonstrate that the defect was recognisable until the distance from the pipe end was around
0.7 m. After that, as shown in Figure 6f, the resolution was gradually reduced until it reaches 0.5 m.
Afterwards, the defect reflection started to superimpose on the pipe end reflection, and it travelled
below the limit of the resolution. Therefore, according to this result, SSP with the proposed parameters
can detect small amplitudes that are close to the dominant amplitude only when the distance between
them is greater than 0.5 m. Note that this is the result for 10 cycles of excitation signal with a defect
size of 10% CSA and a centre frequency of 50 kHz.
Figure 4. Synthesised setup for an eight inch pipe with a wall thickness of 8.179 mm and an outside
diameter of 219.08 mm.
36
Appl. Sci. 2018, 8, 1815
(a) (b)
(c) (d)
Figure 5. Results for the synthesised signal before and after applying SSP ((polarity thresholding &
Polarity thresholding with minimisation)PT & PTM). The defect and the pipe end are located at X = 3 m
and X = 4.5 m from the excitation signal. The defect sizes are (a) 6% cross-section area (CSA); (b) 4%
CSA; (c) 2% CSA; and (d) 1% CSA.
(a) (b)
(c) (d)
Figure 6. Cont.
37
Appl. Sci. 2018, 8, 1815
(e) (f)
(g) (h)
Figure 6. Results for the synthesised signal before and after applying SSP (PT & PTM). The defect
(X = 3.5 m) is moved towards the pipe end (X = 4.5 m) in steps of 0.1 m. The defect distances are X= (a)
3.5 m; (b) 3.6 m; (c) 3.7 m; (d) 3.8 m; (e) 3.9 m; (f) 4 m; (g) 4.1 m; (h) 4.2 m.
Furthermore, as shown in Figure 6, the results of PT and PTM algorithms are almost identical
while the distance between the two peaks is greater than 0.8 m. However, the PT algorithm gives
a better temporal resolution when the distance is around 0.5 m (Figure 5f). In these regions, the PT
identify defects partially, whereas the PTM loses the information completely, and finally, when the
distance is less than 0.4 m, both algorithms miss the defect. The result of the synthesised signal showed
that SSP application has the potential to reduce the level of coherent noise significantly, due to the
presence of DWM, hence enhancing the SNR and the spatial resolution of signals.
In addition, a threshold (0.4 m distance between two features) was defined as a distance limitation,
below which threshold the temporal resolution will be reduced. However, further work is required for
the different scenarios. Furthermore, in order to validate the outcome of these synthesised results, SSP
was applied to the experimental data that was gathered in the lab. These validations are described next.
4. Experimental Validation
In this section, two experiments were carried out in the laboratory to validate the proposed
method for the reduction of DWM, and to enhance the spatial resolution in GWT.
4.1. Experiment #1: One Saw Cut Defect

The first experiments were conducted in the lab using a nominal eight-inch steel pipe, 6 m long,
with a wall thickness of 8.28 mm and an outer diameter of 219.08 mm. The setup for the experiment is
illustrated in Figure 7. The signal was excited/received (Tx/Rx) using a Teletest system to transmit
a 10-cycle Hann window modulated tone burst of T(0,1) wave mode. The ring spacing between the
transducers was 30 mm.
38
Appl. Sci. 2018, 8, 1815
7HOHWHVW&ROODU 6DZ&XWWR&6$
PP
P P P
(a)
5 ring tool
Saw cut defects
0.5% to 8% CSA
Teletest
Compressor
(b) (c)
Figure 7. Experimental setup up for an eight inch steel pipe with a wall thickness of 8.179 mm and an
OD of 219.08 mm (a,b), and (c) its flaw size plan.
The frequencies that give the best results for this particular pipe size according to the dispersion
curve are 27 kHz, 36 kHz, 44 kHz, 64 kHz, and 72 kHz. Therefore, the data was collected at these
frequencies for analysis. The Teletest Collar was placed at 1.5 m away from the near pipe-end, and a saw
cut defect was introduced 1.5 m away from the far pipe-end. The size of the defect was incrementally
increased from 0.5% CSA to 8% CSA in nine steps. The flaw size plan is displayed in Figure 7c. In
order to reduce incoherent noise, the collection was repeated 512 times, and the received signals were
averaged. The sampling frequency of the received signal was set to 1MHz. Since PT algorithm gave the
best result for synthesised signal, it was used in this experiment with the same filter bank parameters
that were employed for the synthesised signal as illustrated in Table 1.
As mentioned earlier, the experiment was run for different frequencies, and the result of each
frequency was compared with the conventional model that is currently employed in the Teletest
system. However, the comparison for 44 kHz is presented in greater detail in this paper. The results
indicated that defects that are smaller than 2% CSA are almost impossible to identify before and/or
after applying the proposed technique. Therefore, the investigation and comparison were confined
to when the defect size was greater than 2% CSA. It is notable that the current sensitivity for reliable
detection of the Teletest system was 9% CSA, which is equivalent to 5% amplitude reflection.
Figure 8 shows the zoom-in plot around the defects area from 0.5% CSA to 8% CSA using MatLab
software for both the unprocessed (blue trace) signal and the SSP (red trace) signal. The figure confirms
that the SSP technique using optimum parameters enhances the defect sensitivity down to 2% CSA,
which was hidden below the noise level. As shown in Figure 8, it was still difficult to identify a
3% CSA defect with the conventional techniques, whereas the proposed method removed all of the
surrounding noise, and only the defect’s reflection remained. Hence, the defect was easily noticeable
and it could be identified with confidence. This size flaw was typical of that which can be challenging
39
Appl. Sci. 2018, 8, 1815
to reliably detect with GWT systems. Therefore, as a result of this experiment, SSP demonstrates great
potential to enhance the SNR, and to increase the sensitivity and spatial resolution of signal response,
and it is able to identify defects down to 2% CSA.
SNR Calculation
In order to quantify the improvements shown by the proposed technique, the SNR enhancement
was calculated as:
S
SNR = 20 × log10 (6)
N
where S is the maximum amplitude of the defect’s reflection, and N is the root mean square (RMS)
value of the noise region around the defect. The SNR of the unprocessed signal was 7.8 dB when
the defect size was 2% CSA, and 13.25 dB when the defect size was 3% CSA. The results presented
in Table 2 show that the SSP algorithms enhanced the SNR by 34.9 dB and 42.9 dB for these cases
respectively. Furthermore, a comparison of the amplitude of the pipe-end reflection and the defect
reflection was undertaken in order to evaluate how well the SSP maintains the amplitude of the signal
of interest. The results indicate that when the size of the defect was 3% CSA or greater, there were no
significant amplitude changes that occurred for the proposed method.
Figure 8. Zoom in around the defect area from 0.5% CSA up to 8% CSA.
Table 2. Signal-to-noise ratio (SNR) enhancement of experimental signal.
Signal 2% Defect 3% Defect

SSP 34.9 dB 42.9 dB
Unprocessed 7.8 dB 13.3 dB
Figure 9 illustrates the result of the unprocessed signal (blue trace) and the SSP algorithm (red
trace) for the above experiment when there was no defect (baseline), up to when the defect size was
8% CSA. The results clearly illustrated that the performance of the proposed technique massively
improved the SNR of the GWT response compared to the unprocessed data, achieving around 30 dB
40
Appl. Sci. 2018, 8, 1815
improvement for 44 kHz. However, as mentioned earlier, the comparison commenced when the defect
size was at least 2% CSA. Hence, in order to clarify it, a dotted line was added at 2% CSA.
N+]
64
54
44
34 aG%
LPSURYHPHQW
dB
24
14
4
-6
0 1 2 3 4 5 6 7 8
CSA (%)
Figure 9. SNR calculation–peak amplitude of the defect (S) to the root mean square (RMS) value of the
noise region (N) for 44 kHz.
4.2. Experiment #2: Two Saw Cut Defects

The aim of this test was to investigate the distance limitation of SSP where a small feature (i.e.,
defect) is close to a dominant feature (e.g., weld). The initial investigation started with the optimum
parameters utilised for a single saw cut defect to evaluate the outcome of SSP distance limitation.
An experiment was carried out on the same eight-inch pipe that was utilised for the previous
experiment. In this experiment, another saw cut defect was added to the pipe at a location that was
48 cm from the far pipe end, as shown in Figure 10. The size of the defect was incrementally increased
from 1% CSA to 8% CSA. Note that another 8% CSA saw cut defect already existed 1.5 m from the
far pipe end. All of the setups, the tool location, and the frequency test region were exactly the same
as the previous exercise. However, only the results relating to 44 kHz were presented here because,
according to the previous experiment, this frequency gave the best SNR enhancement.
7HOHWHVW&ROODU 6HFRQG6DZ&XWWR&6$

)LUVW6DZ&XW&6$
PP
P P P
Figure 10. Experimental setup for the same eight inch pipe (Figure 7) with two saw cut defects.
The proposed method was applied to the collected data gathered from this experiment. It was
observed that the initial result obtained with these parameters was not as successful, and it was only
able to identify defects down to 4% CSA. Hence, the brute force search algorithm was applied to find
the optimum parameter values for this scenario, again in order to improve the performance of the
defects, and the capability to find smaller defects. As a result, optimum values were discovered that
gave the chance to detect defects down to 2% CSA for a second defect. The unprocessed signal, and
the signal after applying SSP with the new optimum parameters are presented in Figure 11. This
figure clearly demonstrates that the data after applying SSP are tidier in general, and in terms of defect
detection, down to 2% CSA is noticeable. However, it can be seen that the coherent noise is hardly
reduced by the new parameters, and SNR is slightly improved. Therefore, it is confirmed that there
is a trade-off between detecting small features next to a dominant signal and improving the SNR. In
41
Appl. Sci. 2018, 8, 1815
addition, this result confirms the result of the synthesised analysis where it was stated that the distance
limitation of SSP to identify adjacent features is around 50 cm when using a 10-cycle Hann windowed
at 44 kHz.
Figure 11. Zoom in result with two defects. (a) Unprocessed signal; (b) SSP signal.
4.3. Discussion
The limitations of SSP were first investigated for synthesised signals using the two most common
SSP recombination algorithms, polarity thresholding (PT) and PT with minimisation (PTM). The
synthesised signals were utilised to find the limitations of SSP by measuring the SNR and spatial
resolution of the received signal. Results showed that the SSP with PT algorithm was able to identify
defects down to 1% CSA, and stated that 50 cm is the minimum distance for the SSP to identify a defect
next to the pipe end. Then, in order to validate the synthesised results, two experimental tests have
been carried out in the lab on the same sized pipe.
Two experiments were carried out on an eight inch pipe, 6 m long, with a wall thickness of
8.28 mm and an outer diameter of 219.08 mm, using a pulse-echo technique. The Teletest system was
utilised to transmit a 10-cycle Hann window-modulated tone burst of T(0,1) with a centre frequency
of 44 kHz. In the first experiment, a saw cut defect was created 1.5 m from the far pipe-end. Nine
saw cut defects were created, the sizes of which gradually increased from 0.5% CSA to 8% CSA. The
results demonstrate that the SNR was improved by approximately 30 dB compared to the unprocessed
signal. The results indicated that defects smaller than 2% CSA cannot be identified both before and
after SSP. This is due to the sensitivity of the system. Therefore, the investigation and comparison were
conducted only when the defect size was greater than 2% CSA. It was shown that the 2% CSA defect’s
reflection was almost masked by the coherent noise level, and that the identification of responses
using conventional signal interpretations was not feasible. However, SSP removed all the surrounding
coherent noise but the defect’s reflection. Therefore, this provided a good evidence that SSP has the
potential to identify defects down to 2% CSA, and to enhance the spatial resolution.
In the second experiment, a new saw cut defect was created, which varied in size from 1% CSA to
8% CSA in each test, in addition to the already existing 8% CSA defect. The aim was to validate the
distance limitation that was observed for the synthesised exercise. Therefore, the defects were created
48 cm from the far pipe-end. The results illustrated that defects up to 4% CSA were detectable with the
same filter bank parameters as those used for the previous scenarios. However, in terms of identifying
smaller defects, the parameters needed to be modified, and this was undertaken using a brute force
search algorithm. As a result, new filter bank parameters were introduced to identify defects down to
2% CSA. However, this was achieved at the cost of reducing the SNR enhancement.
42
Appl. Sci. 2018, 8, 1815
The above algorithm and methodology explain a means of identifying potentially reusable
parameters of SPP for the application of suppressing dispersive wave modes in GWT. It should be
noted that pipes of the similar geometry and material share the same guided wave characteristics.
The relative rates of dispersion between the desired/undesired wave modes will be the same for
equivalent pipes. Regulation in the oil and gas sectors means that pipelines are manufactured from a
limited set of standard geometries and materials. Hence, equivalent pipes are commonplace. Therefore,
it is anticipated that once the set of parameters have been identified for a specific pipe, they can be
reused effectively for similar or slightly different pipes. This would mean it would not be required
to run the brute force search algorithm for every inspection. However, the proposed technique was
tested in a limited trial, and in order to build a signal processing toolbox (automated), more field data
analysis is required to be investigated.
5. Conclusions
A novel solution based on signal processing is proposed in this work to address the problem of
coherent noise in guided wave testing using the SSP technique. The main concern was to identify the
limitations of SSP in terms of sensitivity and resolution when two features are close to each other.
Therefore, a synthesised signal has been created to identify the limitations of SSP in terms of
establishing the smallest defect that can be detected, and to establish the resolution limitation when
the location of a defect is close to a dominant feature. It was demonstrated that the SSP technique with
optimum parameters successfully identified defects down to 2% CSA, and enhanced the SNR of the
received signal. In addition, a threshold of 50 cm has been defined, below which the temporal resolution
will be significantly reduced. The outcome was then experimentally validated for an eight-inch pipe
containing two saw cut defects using the Teletest system, where it achieved a comparable result. To sum
up, it is demonstrated throughout this work that the proposed method using a polarity thresholding
algorithm has the potential to improve the sensitivity and spatial resolution of GWT in terms of SNR
(by up to 30 dB), by detecting smaller defects (down to 2% CSA) with a resolution threshold of 50 cm
between the two features. Thus, this work shows that SSP, as implemented here, could be applied for
pipeline inspection using the GWT technique.
Further work on this subject should focus on validation of this technique with field data, which
are usually more complex and contain different types of defects. Furthermore, the sensitivity of this
algorithm could be investigated for coated and buried pipelines, where the attenuation rates are
sufficiently high that they cause a major reduction in guided wave test capability.
Author Contributions: S.K.P. conducted the synthesised modelling, analysis, programming, experiments, and
drafted the manuscript; P.M. helped edit the manuscript; T.-H.G. contributed to reviewing the manuscript.
Funding: This research was funded by the National Structural Integrity Research Centre (NSIRC), managed by
TWI with the partnership of Brunel University.
Acknowledgments: The authors gratefully acknowledge TWI and PI Ltd. for granting access to the
Teletest equipment.
References
1. Rose, J.L. An introduction to ultrasonic guided waves. In Proceedings of the 4th Middle East NDT Conference
and Exhibition, Manama, Bahrain, 2–5 December 2007.
2. Rose, J.L. Ultrasonic guided waves in structural health monitoring. Key Eng. Mater. 2004, 270–273, 14–21.
[CrossRef]
3. Catton, P. Long Range Ultrasonic Guided Waves for the Quantitative Inspection of Pipelines. Ph.D. Thesis,
Brunel University, Uxbridge, UK, 2009.
4. Fateri, S.; Lowe, P.S.; Engineer, B.; Boulgouris, N.V. Investigation of ultrasonic guided waves interacting with
piezoelectric transducers. IEEE Sens. J. 2015, 15, 4319–4328. [CrossRef]
43
Appl. Sci. 2018, 8, 1815
5. Mallett, R.; Mudge, P.; Gan, T.; Balachandra, W. Analysis of cross-correlation and wavelet de-noising for the
reduction of the effects of dispersion in long-range ultrasonic testing. Insight-Non-Destr. Test. Cond. Monit.
2007, 49, 350–355. [CrossRef]
6. Wilcox, P.; Lowe, M.; Cawley, P. The effect of dispersion on long-range inspection using ultrasonic guided
waves. NDT E Int. 2001, 34, 1–9. [CrossRef]
7. Wilcox, P.D. A rapid signal processing technique to remove the effect of dispersion from guided wave signals.
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2003, 50, 419–427. [CrossRef] [PubMed]
8. Zeng, L.; Lin, J. Chirp-based dispersion pre-compensation for high resolution Lamb wave inspection. NDT E
Int. 2014, 61, 35–44. [CrossRef]
9. Xu, K.; Ta, D.; Moilanen, P.; Wang, W. Mode separation of Lamb waves based on dispersion compensation
method. J. Acoust. Soc. Am. 2012, 131, 714–722. [CrossRef] [PubMed]
10. Xu, K.; Ta, D.; Hu, B.; Laugier, P.; Wang, W. Wideband dispersion reversal of lamb waves. IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 2014, 61, 997–1005. [CrossRef] [PubMed]
11. Xu, K.; Liu, C.; Ta, D. Ultrasonic guided waves dispersion reversal for long bone thickness evaluation:
A simulation study. In Proceedings of the 2013 35th Annual International Conference of the IEEE Engineering
in Medicine and Biology Society (EMBC), Osaka, Japan, 3–7 July 2013; pp. 1930–1933.
12. Toiyama, K.; Hayashi, T. Pulse compression technique considering velocity dispersion of guided wave.
Rev. Prog. Quan. Nondestr. Eval. 2008, 975, 587–593.
13. Yucel, M.; Fateri, S.; Legg, M.; Wilkinson, A.; Kappatos, V.; Selcuk, C.; Gan, T.-H. Coded Waveform Excitation
for High Resolution Ultrasonic Guided Wave Response. IEEE Trans. Ind. Inform. 2016, 12, 257–266. [CrossRef]
14. Yucel, M.; Fateri, S.; Legg, M.; Wilkinson, A.; Kappatos, V.; Selcuk, C.; Gan, T.-H. Pulse-compression based
iterative time-of-flight extraction of dispersed ultrasonic guided waves. In Proceedings of the 2015 IEEE 13th
International Conference on Industrial Informatics (INDIN), Cambridge, UK, 22–24 July 2015; pp. 809–815.
15. Newhouse, V.; Bilgutay, N.; Saniie, J.; Furgason, E. Flaw-to-grain echo enhancement by split-spectrum
processing. Ultrasonics 1982, 20, 59–68. [CrossRef]
16. Karpur, P.; Shankar, P.; Rose, J.; Newhouse, V. Split spectrum processing: Optimizing the processing
parameters using minimization. Ultrasonics 1987, 25, 204–208. [CrossRef]
17. Shankar, P.M.; Karpur, P.; Newhouse, V.L.; Rose, J.L. Split-spectrum processing: Analysis of polarity threshold
algorithm for improvement of signal-to-noise ratio and detectability in ultrasonic signals. Trans. Ultrason.
Ferroelectr. Freq. Control 1989, 36, 101–108. [CrossRef] [PubMed]
18. Saniie, J.; Oruklu, E.; Yoon, S. System-on-chip design for ultrasonic target detection using split-spectrum
processing and neural networks. Trans. Ultrason. Ferroelectr. Freq. Control 2012, 59, 1354–1368. [CrossRef]
[PubMed]
19. Gustafsson, M.G.; Stepinski, T. Adaptive split spectrum processing using a neural network. Res. Nondestruct.
Eval. 1993, 5, 51–70. [CrossRef]
20. Gustafsson, M.G. Towards adaptive split spectrum processing. IEEE Ultrason. Symp. 1995, 1, 729–732.
21. Rubbers, P.; Pritchard, C.J. Complex plane Split Spectrum Processing: An introduction. NDT Net 2003, 8, 1–8.
22. Rubbers, P.; Pritchard, C.J. An overview of Split Spectrum Processing. J. Nondestruct. Test. 2003, 8, 1–11.
23. Rodríguez, A. A new filter bank design for split-spectrum algorithm. NDT 2009, 1, 1–7.
24. Rodriguez, A.; Miralles, R.; Bosch, I.; Vergara, L. New analysis and extensions of split-spectrum processing
algorithms. NDT E Int. 2012, 45, 141–147. [CrossRef]
25. Syam, G.; Sadanandan, G.K. Flaw Detection using Split Spectrum Technique. Int. J. Adv. Res. Electr. Electron.
Instrum. Eng. 2014, 3, 8118–8127.
26. Pedram, S.K.; Fateri, S.; Gan, L.; Haig, A.; Thornicroft, K. Split-Spectrum Processing Technique for SNR
Enhancement of Ultrasonic Guided Wave. Ultrasonics 2018, 83, 48–59. [CrossRef] [PubMed]
27. Eddyfi Ltd. 2018. Available online: http://www.teletestndt.com/ (accessed on 10 July 2018).
28. Beasley, H.R.; Eric, W. A Quantitative Analysis of Sea Clutter Decorrelation with Frequency Agility.
IEEE Trans. Aerosp. Electron. Syst. 1968, 4, 468–473. [CrossRef]
29. Pedram, S.; Haig, A.; Lowe, P.; Thornicroft, K.; Gan, L.; Mudge, P. Split-Spectrum Signal Processing for
Reduction of the Effect of Dispersive Wave Modes in Long-range Ultrasonic Testing. Phys. Procedia 2015, 70,
388–392. [CrossRef]
44
Appl. Sci. 2018, 8, 1815
30. Pavlakovic, B.; Lowe, M.; Alleyne, D.; Cawley, P. Disperse: A general purpose program for creating dispersion
curves. In Review of Progress in Quantitative NDE; Springer: New York, NY, USA, 1997; pp. 185–192.
31. Bottger, W.; Schneider, H.; Weingarten, W. Prototype EMAT system for tube inspection with guided ultrasonic
waves. Nucl. Eng. Des. 1987, 102, 369–376. [CrossRef]
45
applied
sciences
Article
Detection of Defects Using Spatial Variances of
Guided-Wave Modes in Testing of Pipes
Houman Nakhli Mahal 1,2, *, Kai Yang 3 and Asoke K. Nandi 1
1 Department of Electronic and Computer Engineering, Brunel University London, Uxbridge UB8 3PH, UK;
asoke.nandi@brunel.ac.uk
2 NSIRC, Granta Park, Cambridge CB21 6AL, UK
3 TWI, Granta Park, Cambridge CB21 6AL, UK; kai.yang@twi.co.uk
* Correspondence: houman.nakhlimahal@brunel.ac.uk
Received: 19 October 2018; Accepted: 19 November 2018; Published: 24 November 2018
Abstract: In the past decade, guided-wave testing has attracted the attention of the non-destructive
testing industry for pipeline inspections. This technology enables the long-range assessment of
pipelines’ integrity, which significantly reduces the expenditure of testing in terms of cost and time.
Guided-wave testing collars consist of several linearly placed arrays of transducers around the
circumference of the pipe, which are called rings, and can generate unidirectional axisymmetric
elastic waves. The current propagation routine of the device generates a single time-domain signal
by doing a phase-delayed summation of each array element. The segments where the energy of
the signal is above the local noise region are reported as anomalies by the inspectors. Nonetheless,
the main goal of guided-wave inspection is the detection of axisymmetric waves generated by the
features within the pipes. In this paper, instead of processing a single signal obtained from the general
propagation routine, we propose to process signals that are directly obtained from all of the array
elements. We designed an axisymmetric wave detection algorithm, which is validated by laboratory
trials on real-pipe data with two defects on different locations with varying cross-sectional area (CSA)
sizes of 2% and 3% for the first defect, and 4% and 5% for the second defect. The results enabled the
detection of defects with low signal-to-noise ratios (SNR), which were almost buried in the noise
level. These results are reported with regard to the three different developed methods with varying
excitation frequencies of 30 kHz, 34 kHz, and 37 kHz. The tests demonstrated the advantage of using
the information received from all of the elements rather than a single signal.
Keywords: signal processing; defect detection; spatial domain; array analysis; pipeline inspection;
ultrasonic guided waves (UGWs)
1. Introduction
Pipelines are the main means of transferring oil and gas. They are usually installed in a hostile
environment and suffer from defects such as corrosion or mechanical damage. To prevent the failure
of pipelines, there is a need to inspect them in service in order to ensure their structural integrity.
One of the many methods for the non-destructive testing of pipes is ultrasonic guided waves (UGW).
This technology uses sets of transducer arrays to generate a low-frequency ultrasound signal. This
ensures that elastic waves are guided through the boundary of the pipe; hence, the name guided
waves [1]. Therefore, from one test point, the long-range inspection of pipelines is possible. However,
the interoperation of UGW signals tends to be difficult due to the existence of the coherent noise in
the tests.
Despite using arrays of transducers, UGW testing systems use a general routine process that
outputs a single monodirectional signal in a time domain [2–5]. The generated signal is then used
by inspectors to report any anomalies above the regional signal noise level [6]. To allow an ease of

Appl. Sci. 2018, 8, 2378
inspection, many types of research have been defined to use the characteristics of guided waves to
increase the signal-to-noise ratio (SNR), increasing the defect detection accuracy.
One of the widely used techniques in the signal processing of guided waves is the work of
Wilcox et al., which uses previously calculated dispersion curves in order to compensate for the
effect of dispersion for the wave mode of interest in a certain propagation distance [7]. In 2013,
Zeng et al. [8] used the basis of the dispersion compensation technique and introduced a novel method
to design waveforms in order to pre-compensate for a certain propagation distance. Various researchers
investigated the pulse compression approach, which consisted of exciting a known coded sequence
and match-filtering the results in order to detect the known sequence [9–12]. Mehmet et al. [9,10]
introduced an iterative dispersion compensation for removing the dispersion effect of guided-wave
propagation. In 2016, Lin et al. [11] reported on the performance of different excitation sequences
in terms of their ability to achieve δ-like correlation on Lamb waves. Most recently, Malo et al. [12]
developed a two-dimensional compressed pulse analysis in order to enhance the achieved SNR of
each wave mode. Coded waveform approaches have also been implemented in the enhancement of
SNR in the inspection of Long Bones using guided waves [13]. In the pulse compression approaches,
the results are always a trade-off between the spatial resolution and SNR; for having a high-energy,
δ-like correlation result, the wavelength must be increased. Furthermore, the transfer function of
the excitation system can affect the accuracy of the coded waveforms. These two factors mean that
the initial waveform design and the accuracy of inspection using this method depend on the testing
conditions. Furthermore, if non-dispersive regions of wave modes are used, there will be no need to
use dispersion compensation algorithms.
Another method for filtering the guided waves signal is the work of Pedram et al. in
Split-Spectrum Processing (SSP) [14,15], whereby through selecting the optimum filter parameters,
the coherent noise of the test cases can be removed. However, automatic selection of the optimum
filtering parameter is not possible at present. These parameters can be varying with regard to central
excitation frequencies and the pipe geometries. On a similar note, frequency sweep examination, which
was developed by Fateri et al. [16] in 2014, uses a similar concept in a way that defect features should
exist in all of the frequency bands. Based on this fact, a novel method was developed that processes
signals with different excitation frequencies, and is able to distinguish between the dispersive and
non-dispersive wave modes that exist in the test. This method has never been tried in the pipes before,
and requires many different excitations to be calculated on the same test subject. As it is currently
known, due to backward cancellation algorithms and an overall system transfer function [4], it is not
perfect, and not all of the frequencies can be used for optimum excitation (it can be varying depending
on the pipe and tool set used). This will affect the accuracy of the algorithm if it is converted for
pipeline inspection.
In 2015, Ching-Tin Ng [17] investigated signal processing approaches for the enhancement of
results produced by guided-waves signals, in which he reported that Hilbert transform, Gabor wavelet
transform, and discrete wavelet decomposition are able to improve the detectability of the results.
UGWs have also been used for monitoring purposes. In 2015, Dobson et al. implemented independent
component analysis (ICA) combined with previously developed temperature compensation
methods [18,19] to enable the successful guided-wave monitoring of pipelines. In 2017, Chang et al. [20]
introduced the idea of an adaptive sparse deconvolution method, with the aim of distinguishing the
overlapping echoes of an ultrasonic guided-waves signal.
All of the algorithms have been developed with one purpose in mind: the improvement of the
SNR to increase the possibility of the detection of a defect signal. However, the full potential of
the system has not yet been established. The signal information in these systems are received from
three rings of (at least) 24 transducers, but in the mentioned work, there is only one signal, which is
achieved after the propagation routine of the device by summing the phase-delayed version of all of
the transducers together. This routine was created to reduce the received flexural signals and have
a unidirectional inspection [4,5]. Nonetheless, the received informatics can be processed in a better
48
Appl. Sci. 2018, 8, 2378
method, i.e., to use all of the available sensors for detecting the defect location rather than changing
the information into a single time-domain signal.
On the other hand, array processing techniques that use individual signals are generally used for
imaging and focusing applications [21–27]. However, these algorithms are computationally expensive,
and might require the individual excitation of transducers [26], or a phased-array focusing of the
transducers [23], rather than synthetic focusing, where all of the transducers in the array are excited at
the same time [21]. This adds to the complexity of the interpreting the results. Furthermore, the goal of
these algorithms is imaging a segment of the pipe rather than the detection of the defect location.
In this paper, we propose a new algorithm to solve the problem of detections, which can possibly
be combined with all of the above-mentioned methods for SNR enhancement. The aim of this algorithm
is to illustrate the possibility of using all of the individual transducers for defect detection instead
of using a single signal achieved after the general routine. This way, both the spatial and temporal
characteristics of the signal can be examined in more depth, which can lead to more accurate results.
Furthermore, unlike imaging techniques, it uses normal excitation sequences, and thus can be applied
to all of the data sequences, even if they are achieved in the past.
Section 2 explains the nature of noise on the guided-waves signal and the wave formation with
regards to unique features. In Section 3, a novel method is proposed that uses the characteristic of
axisymmetric waves to allow the detection of the defect signal. Validation was done by laboratory
trials using two separate defect locations. The first defect is in a region where less coherent noise
exists in the baseline with a varying size of 2% and 3% of the pipe’s cross-sectional area (CSA), and the
second defect is located in higher spatial noise with a varying size of 4% and 5% CSA. The setup of
these tests and the results are reported in sections 4 and 5, respectively. In the last section, a summary
of current achievements as well as the further works is reported.
2. Background Theory
Guided waves are dispersive and multimodal [2]. Dispersion causes the energy of a signal to
spread out in space and time as it propagates [7]. Guided wave modes are categorized based on their
displacement patterns (mode shapes) within the structure. Three main families of waves exist in pipes,
which are longitudinal, torsional, and flexural. Longitudinal and torsional waves are axisymmetric
waves, while flexural waves are non-axisymmetric. The popular nomenclature used for them are in the
format of X(n,m), where X can be replaced by the letters L for longitudinal, T for torsional, and F for
flexural waves; n shows the harmonic variations of displacement and stress around the circumference,
and m represents the order of existence of the wave mode [28].
Due to the sinusoidal and spatial variations of flexural waves, these coherent noises in the
inspection can be reduced by adding the signals received from linearly spaced transducers around
the circumference of the pipe. The resultant effect is the cancellation of non-axially symmetric
flexural waves and the enhancement of axially symmetric torsional or longitudinal waves. However,
cancellation is not ideal because of the imperfections of the real-world devices. These factors include the
non-linear spacing of the transducer’s placement, the variability of the transducer’s frequency response,
the number of transducers used, or the tolerances of the electronic test device. Furthermore, a scattering
of an ideal axisymmetric wave with a non-axisymmetric feature would also result in the generation
of other wave modes, which is commonly known as a wave mode conversion phenomenon [29–32].
Thus, inspecting guided waves without the interference of these coherent noises is near impossible.
For decades, the inspectors have been trained to interpret these resultant time-domain signals to
extract the defect location. Nevertheless, the problem of detection can be defined in another way, i.e.,
the linearity and symmetry between the phases of the received waveforms. For better understating of
this matter, in the following section, different conditions of waves arrived at spatially are demonstrated
using a synthesized model.
49
Appl. Sci. 2018, 8, 2378
2.1. Wave Formation

The scattered wave would consist of the excited axisymmetric and its associated flexural waves.
Depending on the feature shapes, the energy distribution of each of the created wave modes would be
different. As an example, a perfectly axisymmetric feature in the ideal scenario would result in the
generation of a perfect axisymmetric response with no other flexural waves. In these scenarios, if the
axisymmetric wave is stronger than the coherent regional noises in the area, an anomaly is reported by
the inspectors. However, since it is a visual interpretation, in some scenarios, the resultant mixture
of flexurals can easily lead to false alarms due to their coherence to the expected received signal.
Furthermore, the axisymmetric wave can be smaller than the local noise level, and therefore would not
be detected. These are some of the main challenges faced when inspecting the temporal domain.
Figure 1 shows the spatial arrival of the signals received from two rings of 32 source points. These
signals are generated by the finite element model (FEM) [30]. In this setup, ideal source points are
assumed, which leads to the generation of a perfect T(0, 1) wave mode in all of the figures except (b)
and (i), where the frequency response of each transducer is variable.
(a) Pipe End: Pure T(0,1) Wave (b) Pipe End: Impure T(0,1)) (c) Pipe End: T(0,1) after Defect
(d) Defect: Start of Defect (e) Defect: Near the peak (f) Defect: Near the end
(g) Defect: Located at 0° ((h) Noise: Flexurals caused by defectt (i) Backward Leakage
Figure 1. Spatial signal reception from two rings of 32 source points from various features where the
red, blue, and black dotted lines are showing the first ring, second ring, and the reference offset.
50
Appl. Sci. 2018, 8, 2378
The red and blue signals are showing the first and second ring, respectively, and the reference is
illustrated by the black dotted line. Thus, the values inside the reference would mean a negative phase,
and values higher than the circle would mean a positive amplitude of the individual transducer.
2.1.1. Pipe End

The received signals from the pipe end are shown in Figure 1a–c. As can be seen in (a), a pure
T(0, 1) wave is generated, since both the excitation sequence and the feature are perfectly axisymmetric.
However, in the cases of (b) and (c), where the flexurals are created by imperfect excitation and from
wave mode conversion, respectively, the received response is the superposition of T(0, 1) and flexural
waves. The scattering from the pipe end would result in all of the energy of the wave to be reflected.
Thus, a strong proportion of the received signal energy corresponds to the T(0, 1) wave, and small
variations are caused by flexural waves.
2.1.2. Defect
In cases (d)–(g) of Figure 1, the received waves is generated from a non-axisymmetric defect with
a cross-sectional area size of 3%. The shown signals are different sequences from various times with
respect to the temporal domain signal. The capture times are shown in Figure 2b where the red line is
representing case (d), the blue line is representing case (e), and the green line is representing case (f).
Case (g) is captured at the same time as case (e), but the generated wave is from a defect located at
0◦ (the right side of the pipe), while the other cases were from a defect angled at 90◦ (top side). Four
important points are illustrated from the figures:
1. Unlike the pipe end, defects that are non-axisymmetric features will inherently generate flexural
waves. Nonetheless, the strongest energy will be associated with the T(0,1) wave.
2. The end of the signal envelope will contain more flexural waves compared to the start of the
envelope. At the start of the envelope, at most one cyclic variation will exist in the spatial
domain, which would represent a low-order flexural wave mode. Nonetheless, near the end of
the envelope, higher order variations (three or more) are starting to increase. This is because after
the center of the envelope, the power of the T(0,1) wave is starting to decrease, and the higher
order flexural with a lower travel speed is beginning to arrive.
3. The strongest energy of the T(0,1) wave is associated with the peak of each cycle in the temporal
domain, where most of the transducers will have the same phase.
4. When the angle of the defect is changed, the direction of the arrival changes, but the phases and
the overall formation of the wave remain the same.
Start of Near Peak
Defect
End of Defect
(a) Backward Leakage (b) Defect (c) Flexurals
Figure 2. Temporal domain of cases (d)–(i) from Figure 1
51
Appl. Sci. 2018, 8, 2378
2.1.3. Coherent Noise

The spatial arrival of the flexural noises created by the wave mode conversion and the backward
leakage due to the imperfect excitation is shown in Figure 1h,i, respectively. Their corresponding time
domain signals are shown in Figure 2b,c. As can be seen, for the flexural signal, the reception point
phases are highly non-linear, which is completely different from the case of the defect signal.
Backward cancellation is not always perfect, due to the fixed ring spacing and the variance along
the received waves from each transducer. In the temporal domain, these waves might still be detected,
but when one looks at the phases between the two rings, it is observable that the two rings would
have the exact opposite phases. Thus, it would be easier to detect signals received from the opposite
direction in the spatial domain.
3. Methodology
In the previous section, it was demonstrated that in the case of flexural waves, high-order
sinusoidal variation along the pipe’s circumference will be created. In ideal scenarios, where a perfect
axisymmetric wave hits a perfect axisymmetric feature, the received amplitude in each transducer
must be the same. However, due to the imperfections of the transducers and the testing conditions,
there will be variance along each of the received amplitudes from the transducers; nonetheless, they
all will have the same phase. Similar to the pipe ends, defects would generate a significant amount of
torsional mode. However, they also do create flexurals with respect to the defect size and location. For
signals coming from the backward direction, it is observed that the phases would practically be the
inverse of each other for each transducer.
Based on these observations, by detecting the linearity between the phases of the sensors,
the axisymmetric waves can be detected. Since not all of the phases would become the same in
the case of defect signals, a measure of similarity between the phases must be defined, which can then
be the threshold for detection.
3.1. General Routine

The first step in the general propagation routine of guided wave devices is to cancel the signal
received from the backward direction. For doing so, the signals received from Ring 2 are initially
subtracted from Ring 1 and Ring 3, and then the results are phase-delayed with respect to the tool set’s
ring distancing and the excitation wave mode dispersion curve (T(0, 1) in this case). The results are
stored in the Signals matrix:

PhaseDelayed(Ring1 − Ring2), for i ≤ NS/2
Signals[i, n] = (1)
PhaseDelayed(Ring3 − Ring2), for i > NS/2
where n represents the sample number, i is the sensor number, and NS represents the total number of
achieved signals, which is twice the number of transducers in each ring. This method will result in
T(0, 1) components to be overlapped. For the signals received from the forward test direction, this
overlap will be positive, and for the backward direction, the amplitudes are inverted, which allows
the unidirectional inspection. The temporal domain signal, which is generally used for inspection, is
achieved by taking the mean of the Signals:
1 NS
NS ∑i = 1
temporal[n] = Signals[i, n] (2)
3.2. Initialization
All of the transducers have the same phase when the sum of all of the signals from the same-phase
transducers is equal to the sum of all of the signals. Therefore, the measure of similarity of the phases
is defined by taking the difference between the sum of all of the signals and the sum of in-phase signals
52
Appl. Sci. 2018, 8, 2378
received from each transducer. The sign of the temporal domain demonstrates which sign is the greater
power of iteration. Based on this sign, a mask is created for each transducer as follows:

1, sign(Signals [i, n]) = sign(temporal [n])
mask[i, n] = . (3)
0, else
Using this mask, in-phase signals with the temporal domain of each sample can be identified,
and their sum can be calculated as:

NS
inPhase[n] = ∑i = 1 mask[i, n]∗Signals[i, n] (4)
Afterward, the similarity can be defined using:

NS
comparisonValue[n] = inPhase[n] − ∑i = 1 Signals[i, n] (5)
3.3. Threshold Selection

In cases where a significant proportion of iterations energy is from the T(0, 1) wavemode,
the comparisonValue will be zero. However, as shown previously, defect signals will contain a
mixture of flexural and torsional wave modes, leading this value to be non-zero. Hence, there is a need
for defining a threshold for detecting the defects. In this section, three methods are presented defining
this threshold.
3.3.1. Method One

In this method, a static constant is defined by the inspector. If the comparisonValue is less than
this defined threshold, the sample would be marked as one, which means that a feature is detected;
otherwise, it would be marked as zero:

1, comparisonValue[n] ≤ ThresholdOne
Result[n] = (6)
0, else
For small signals, this formula would not be accurate, since inherently, their sum would be bigger
than the threshold; thus, a compensation factor, 0.001, is added to prevent outliers from being detected
in the noise region. This is an acceptable level, since the defect signals smaller than 0.001 would not be
detected by the analog to digital converters of the device due to their limited discretization range. If the
amplitude of the signal in the temporal domain is greater than this compensation factor, the algorithm
would work as normal, else it would be zero:

Result[n], |temporal[n]| > compensation
Result’[n] = (7)
0, else
The disadvantages of this method are that firstly, a user input is required, and secondly, the value
is fixed for all of the features, disregarding the existing characteristics in each sample.
3.3.2. Method Two

To use signal characteristics more effectively, the amplitude of the inPhase signals can be used to
define the threshold, since a greater inPhase signal can be the result of a greater linearity of the phases.
By taking a percentage of the inPhase signals, a higher threshold can be assigned to such cases. In this
method, a fixed percentage value needs to be set by the inspector, and the threshold for each iteration
would be adaptively changed:
ThresholdTwo,V1 [n] = inPhase[n]∗percentage (8)
53
Appl. Sci. 2018, 8, 2378
Another improvement to this algorithm would be to consider the effect of comparisonValue as

well. A higher comparisonValue means that there is a higher chance that the signal is not linearly
received by all of the transducers. This effect is significant in the case of signals received from the
backward direction. In each iteration, the comparisonValue can be subtracted from the inPhase, which
would decrease the resultant threshold value for backward propagating or non-axisymmetric waves:
ThresholdTwo,V2 [n] = (inPhase[n] − comparisonValue[n])∗percentage (9)
3.3.3. Method Three

To eliminate the human error, the requested input from the user must be eliminated. The phases
are the main factor in detecting the results; the percentage can be defined based on the ratio of the
in-phase sensors over the total number of sensors:
1 NS
NS i∑
percentage[n] = mask[i, n] (10)
=1
In no iteration is the value of mask zero; it always have a positive offset, which means that the
percentagecannot be replaced directly into Equation (8). Nonetheless, this offset can be removed by
subtracting the mean of the calculated percentage for all the samples from each individual sample:
1 N
N n∑
Average = percentage[n] (11)
=1
where N is the total number of samples in the test. The final considered percentage for each sample
would be:
percentageβ [n] = percentage[n]− Average (12)
Doing so removes the need for user input, and provides a more accurate thresholding with regard
to the phases of the signals in each sample; the higher number of the same phase signals means a
higher percentage value for the sample, which leads to a higher chance of detection. The formula for
the first version (V1) and second version (V2) of this method can be achieved by substituting the new
percentage achieved from Equation (12) in Equation (8) or Equation (9):
ThresholdThree,V1 [n] = inPhase[n] ∗ (percentage [n] − Average) (13)
ThresholdThree,V2 [n] = (inPhase[n] − comparisonValue[n])∗(percentage [n] − Average) (14)
3.4. Limits Definitions

Three main detection limits exist in these algorithms. (1) The first if the minimum, which is the
minimum value required for each method to detect at least one cycle of the defect signal. (2) Second is
the maximum, which is the maximum allowable value for the methods to detect the defect without any
outliers. By moving toward the maximum value, the detection length increases; however, the accuracy
reduces as the chance of detecting outliers increases. Thus, for comparing the methods together,
an average value is defined using the minimum and maximum. This value is the best possible
compromise between the achieved defect length and the possibility of detecting false alarms. For
method three, the average value is calculated using Equation (11), and minimum and maximum
values are achieved by manually replacing the average in Equation (14). For both of the other methods,
the average value is calculated manually from the achieved minimum and maximum values in the
tests. Also, the Sa f e zone margins are defined as the difference between the minimum and maximum,
which illustrate the values where manual inputs will produce a result without any false alarms.
54
Appl. Sci. 2018, 8, 2378
3.5. Program Flowchart

The program flowchart is shown in Figure 3. As can be seen, the three stages of general routine,
initialization, and result generation are the same in all of the methods. However, the threshold value
that is used in the result stage is selected based on methods one, two, or three, which are explained in
the thresholding section.
Figure 3. System flowchart.
4. Experimental Setup
The Teletest device [33], manufactured by Teletest branch of Eddify technology family in
Cambridge (UK), used in order to inspect an eight-inch schedule 40 steel pipe with a length of
six meters. The test tool included three rings of 24 linearly spaced transducers with 30-mm spacing,
which are located 1.5 m away from the back end of the pipe. The main aim of these trials is to assess
the ability of the algorithm in highly noisy scenarios and investigate the dependency of each method
to the excitation frequency.
In the first set of tests, two defects of 3% and 2% CSA have been created; these are located 4.5 m
from the back end of the pipe, which generates signals with less/almost the same energy as the noise
regions. In order to confirm the ability of this thresholding method in detecting another defect in
55
Appl. Sci. 2018, 8, 2378
the same pipe, defects with CSA sizes of 4% and 5% have been created after the initial trials, which
are located 5.5 m away from the back end of the pipe. Figure 4 shows the signal reception routes of
this setup.
)RUZDUG
'HIHFW
%DFNZDUG
'HIHFW
3LSH(QG
7HVW7RRO
3LSH(QG
Figure 4. Signal reception routes.
The excitation sequence is a 10 cycle Hann-windowed sine waves with frequencies of 30 kHz,
34 kHz, and 37 kHz. The selected frequencies are considered the optimum excitation sequences
with regard to their resultant forward propagating energy achieved using the backward cancellation
algorithm [30]. A summary of the settings used for the first set of tests (first defect), which are
numbered from one to six, are shown in Table 1; for the second set of tests, they are numbered from
seven to 12 (second defect), and are shown in Table 2.
Table 1. Summary of the first set of tests conditions (first defect). CSA: cross-sectional area.
Test (#) Defect CSA (%) Frequency (kHz)

1 30
2 3 34
3 37
4 30
5 2 34
6 37
Table 2. Summary of the second set of tests conditions (second defect).
Test (#) Defect CSA (%) Frequency (kHz)

7 30
8 5 34
9 37
10 30
11 4 34
12 37
Figure 5 shows the baseline signals achieved from the same pipe with no defects, with excitation
frequencies of (a) 30 kHz; (b) 34 kHz; and (c) 37 kHz. As can be seen, even though there are no defects
in the pipe, the baseline signal is highly polluted by the coherent noise. The first defect signal is
expected to be received approximately three meters away from the tool test, between the first and
second red dotted lines. The second defect is expected to be located approximately four meters away,
between the third and fourth red dotted lines. Three major conclusions can be drawn with regard to
the baselines:
1. Defect 1: The signal with lowest amount of existing coherent noise in the baseline are 34 kHz,
30 kHz, and 37 kHz, respectively.
2. Defect 2: The lowest amount of existing coherent noise in the baseline belongs to 37-kHz, 34-kHz,
and 30-kHz signals, respectively
56
Appl. Sci. 2018, 8, 2378
3. All of the respective frequencies have a higher amount of noise in the Defect-2 region.
(a) 30 kHz
(b) 34 kHz
(c) 37 kHz
Figure 5. Baseline signals of pipe with no defect with frequencies of (a) 30 kHz; (b) 34 kHz; and
(c) 37 kHz. The first marked region (3–3.5 m) is the location of the first defect, and the second marked
region (4.1–4.5) is the location of the second defect.
It is expected that the thresholds are required for detecting regions with lower existing coherent
noises, which means that overall, Defect 2 will be more difficult to detect.
57
Appl. Sci. 2018, 8, 2378
5. Results
In this section, the limits achieved for detecting the defect signal using each method is reported.
The SNR of the defects in each test is demonstrated in Table 1. These values are generated using the
following equation:
energy(Signal)
SNR = 20 log10 (15)
energy(Noise)
The SNRs are calculated with respect to the overall noises in the region. In the case of test numbers
1–6, only the energy of Defect 1 and the pipe end signal is removed for calculating the energy of the
noise; however, for test numbers 7–12, as well as the Defect-1 and pipe end signals, the Defect-2 signal
is also removed for calculating the noise. As demonstrated in Table 3, the defect signals are within the
noise level, and identification of them using the inspection routine would not be accurate, especially
for a 2% CSA defect. Nonetheless, the overall SNR of Defect 2 is much higher due to both the increment
of CSA size, and the existing coherent noises of the region.
Table 3. Calculated signal-to-noise ratio (SNR) with respect to the coherent noises in the inspection
(in dB).
Test # 1 2 3 4 5 6 7 8 9 10 11 12
SNR 2.97 8.93 10.36 −2.65 −0.98 −3.43 18.57 27.63 28.05 14.73 24.00 23.10
5.1. Example Test Case

Figure 6 shows the Teletest signal of Test 5 before processing, where the defect’s signal region is
in between the two dotted red lines. The SNR of a defect in this test is −0.98 dB, which means that it is
completely buried in the noise. The noise signal is calculated by removing the torsional waves (pipe
end and defect signal) from the shown signal in Figure 6. However, since Defect 2 is not present in this
test, the location in between 4–4.5 m is also considered as noise.
Figure 6. Example signal achieved using general routine, tested by 34 kHz where the first two red
dotted lines shows the defect region, and the third and fourth ones show the pipe end.
Figure 7 shows the assigned values to the first (black line) and the second version (red line)
of thresholding, which is the inPhase and subtraction of the comparisonValue from the inPhase,
respectively. As can be seen both in the backward leakage signal and the noise region, the decrease in
58
Appl. Sci. 2018, 8, 2378
the assigned threshold is more than the defect and pipe end signals. In method two, a static percentage
of either of the mentioned values is considered the threshold for each iteration; however, in method
three, this percentage is assigned using the ratios of inPhase signals subtracted from its mean, as
shown in Figure 8. The calculated threshold for each iteration using the average value of method three
is shown in Figure 9. In the same figure, if method one was used, the threshold value would be a
constant line for all of the iterations. The processed signal calculated from method three is illustrated
in Figure 10, where it shows the locations at which the comparisonValue is higher than the threshold
history. As expected, the pipe end is completely detected, since it is a complete axisymmetric feature,
but only the center part of the defect could be detected.
Defect, Noise, Pipe End
Backward leakage
Figure 7. Comparison of the first and second thresholding versions on the Test 5 signal where the two
zoomed regions are showing the backward leakage, defect, noise, and pipe end, respectively. The black
and red signals are showing the first and second versions, respectively.
Amplitude (arbt.)
Distance (m)
Figure 8. Calculated percentage of sensors with same phase, where the offset (mean values) are
removed (orange); the black signal is showing the filtered version for better illustration.
59
Appl. Sci. 2018, 8, 2378
Amplitude (arbt.)
Distance (m)
Figure 9. Calculated threshold using method three on Test 5 (red) against the second version
comparisonValue (black). In the case of method one, the threshold would be a constant line instead of
changing based on the iteration.
Figure 10. Method three result (green signal) using the average value against the temporal signal
(black) for testing a 2% CSA defect.
5.2. First Set of Tests

The detection of Defect 1 is the focus in the first set of tests (test numbers 1–6). In these test cases,
Defect 2 does not exist, and therefore, the region between 4–4.5 is considered noise. Furthermore,
the coherent flexural noises existing in the Defect-1 region are less than those in the second Defect-2
region, with 34 kHz, 30 kHz, and 37 kHz having the least amounts, respectively. As can be seen from
Table 3, the defects in these tests have a low SNR and are almost buried in temporal noise, which
makes them difficult to be detected in the temporal domain.
5.2.1. Method One

This method was only successful for tests 1, 2, and 4. Figure 11 shows the limits of the threshold
values. Observing the trends with regards to the frequency in tests 1–3, the minimum values are
60
Appl. Sci. 2018, 8, 2378
increasing, while the maximum values are decreasing up until the point where the minimum will
result in an outlier in Test 3. The same trend is observable for the minimum values in tests 4–6, where,
although the maximum values could not be detected, the minimum values are increasing with regard
to the frequency. With regard to the defect CSA (and SNR), the algorithm can detect the 2% CSA defect
using a 30-kHz excitation frequency (Test 4). Nonetheless, when the frequency is increased to 34 kHz
in Test 5, although the defect’s SNR is increased by 1.68 dB, this method resulted in a false alarm, even
at the minimum limit. Thus, it can be concluded that this algorithm is more sensitive to the frequency
and local amplitude of the defect region rather than its temporal SNR and the overall energy of the
noises in the region. Since higher energy is associated with lower frequencies, lower frequencies tests
would achieve better results using this method for Defect 1.
Minimum Average Maximum

0.018
0.016
0.016
Amplitude (arbt.)
0.014
0.0133 0.013
0.012 0.012 0.0125
0.01 0.01
0.008 0.008
0.007
0.006
0.005
0.004 0.004
0.002
0.001 0.001
0
30kHz-3% 34kHz-3% 37kHz-3% 30kHz-2% 34kHz-2% 37kHz-2%
Figure 11. Minimum, average, and maximum values of method one of the first set of tests.
5.2.2. Method Two

This method is effective in all of the tests except for Test 6. With regard to Figure 12, which
shows the limits achieved using this method, the relation between the detection limits and frequency
is no longer linear. The minimum is increasing with the frequency until it reaches a maximum peak
on 34 kHz; then, it starts to decrease when moving toward 37 kHz. The same trend is observable
for the detection length and safe zone margins. This means that this algorithm is a function of both
the SNR of the noise and the defect, as well as the frequency where the optimum frequency would
be around 34 kHz. Nonetheless, it should also be pointed out that when using a lower frequency,
the same happens as in the previous method: the minimum value for detection is closer to zero, which
means that defects would be easier to detect. Regarding the second thresholding version, the increase
in the minimum limits is not as much as those in the maximum, which therefore increases the safe
zone margins in all of the tests. This is clearly observable in tests 1 and 2, where the minimum limits
remained unchanged. The average and maximum detection lengths of the defect remain unchanged.
Therefore, using the second threshold version provides a greater safe zone margin.

0.3 0.45
0.281 0.4 0.39
Amplitude (arbt.)
0.25
0.231 0.234 0.35
0.2 0.196 0.3 0.3 0.305
0.176 0.25
0.159 0.246
0.15 0.146 0.213
0.129 0.2 0.188
0.116 0.163 0.172
0.1 0.15
0.131
0.074 0.073 0.1
0.05 0.08 0.079
0.039 0.039 0.05
0.027 0.04 0.042
0.006 0.027
0 0.001 0 0.001 0.006
30kHz-3%
30kHz 3% 34kHz-3%
34kHz 3% 37kHz-3%
37kHz 3% 30kHz-2%
30kHz 2% 34kHz-2%
34kHz 2% 37kHz-2%
37kHz 2% 30kHz-3%
30kHz 3% 34kHz-3%
34kHz 3% 37kHz-3%
37kHz 3% 30kHz-2%
30kHz 2% 34kHz-2%
34kHz 2% 37kHz-2%
37kHz 2%
(a) V1 Thresholding (b) V2 Thresholding
Figure 12. Limits achieved using method two on the first set of tests.
61
Appl. Sci. 2018, 8, 2378
In comparison to method one, this method has a greater range of operating frequencies and
provides a larger safety margin for threshold selection. The optimum excitation frequency for this
method is 34 kHz for the first set of tests.
5.2.3. Method Three

The average values are calculated automatically by taking the mean of the percentage of active
sensors in each iteration; thus, they are fixed in both thresholding versions. The same as method two,
this method can detect the defects without any false alarms in tests 1–5. However, only tests with an
excitation frequency of 34 kHz can use the average value, since using it for other excitation frequencies
would result in detecting outliers. The limits of this method are shown in Figure 13. The tests where
the average value is between the minimum and maximum are when the automatic detection method
is successful. Furthermore, the averages values achieved using all of the different frequencies are
between 0.602–0.554, and tend to decrease slightly when the frequency is increased.

1.1 1.1
1 0.999 1 0.999
Amplitude (arbt.)
0.9 0.869 0.9 0.869

0.838 0.843 0.859 0.827 0.842
0.8 0.8 0.852
0.753 0.736
0.7 0.646 0.7
0.637 0.5535 0.6 0.599 0.607 0.5535
0.6 0.5905 0.602 0.5905 0.602
0.5639 0.5609 0.5663 0.5639
0.5 0.5 0.5609 0.5663
0.469 0.498
0.4 0.465 0.4 0.409 0.413
0.36
0.3 0.3
30kHz-3% 34kHz-3% 37kHz-3% 30kHz-2% 34kHz-2% 37kHz-2% 30kHz-3% 34kHz-3% 37kHz-3% 30kHz-2% 34kHz-2% 37kHz-2%
Figure 13. Limits achieved using method three in the first set of tests.
To conclude, this method can automatically detect defects of up to 2% CSA, which are buried in
the noise level using a 34-kHz excitation frequency. Furthermore, with regard to the detection length,
it can also be said that 34 kHz is the optimum excitation frequency in the first set of tests.
5.2.4. Comparison
Figure 14 shows the amount of increase in the minimum and maximum values achieved using the
first and second version of the thresholding. The minimums tend to increase less than the maximum
values by at least a factor of two.
Increment of Minimum Incerement of Maximum

0.105
0.109
0.085
Amplitude (arbt.)
0.071
0.069
0.06
0.039
0.038
0.037
0.029
0.017
0.015
0.011
0.007
0.006
0.006
0.001
0
30K HZ - 3% 34K HZ - 3% 37K HZ - 3% 30K HZ - 2% 34K HZ - 2 % 37K HZ - 2% 30K HZ - 3% 34K HZ - 3% 37K HZ - 3% 30K HZ - 2% 34K HZ - 2% 37K HZ - 2%
(a) Method Two (b) Method Three
Figure 14. The increments achieved using the second version of thresholding in the first set of tests.
62
Appl. Sci. 2018, 8, 2378
The generated safe zone margins for all of the methods are shown in Figure 15. Except for Test 4,
method three tends to have a higher safe zone margin than method two, and the best margin is achieved
using second thresholding version for method three. These results are especially important on tests
2 and 5, where the optimum excitation frequency of both methods is used. The second version of
thresholding always produces a greater safety margin.
Method One Method Two (V1) Method Two (V2) Method Three (V1) Method Three (V2)
0.433
0.374
Amplitude (arbt.)
0.273
0.204
0.194
0.172
0.161
0.148
0.137
0.117
0.116
0.082
0.079
0.075
0.073
0.073
0.067
0.06
0.0123
0.017
0.009
0.009
0.01
0
0
30KHZ -3% 34KHZ -3% 37KHZ -3% 30KHZ -2% 34KHZ-2%
Figure 15. Bar graphs of the Safe zone margins of all methods in the first set of tests.
Figure 16 shows the length of detection achieved by average values with different algorithms.
Cases where outliers are detected are not reported. As can be seen, in the optimum frequency, both of
the defect lengths achieved by method two and method three are the same. In Test 5, in the second
version of thresholding, the length has decreased, since the average value is fixed, and the defect size is
smaller; however, in Test 2, the same detection length is achieved using both versions, since the defect
is wider. Method three is the best out of these three mentioned ones, since it produces a larger safe zone
margin, and can detect the defects automatically using the optimum excitation frequency of 34 kHz for
the first set of tests.
0.2
Amplitude (arbt.)
0.155
0.155
0.153
0.151
0.126
0.093
0.093
0.088
0.088
0.09
0.079
0.079
0.075
0.034
0.034
0.026
30KHZ -3% 34KHZ-3% 37KHZ-3% 30KHZ -2% 34KHZ -2%
Figure 16. Bar graphs of the detection length of the defect using the average value of all of the methods
in the first set of tests.
5.3. Second Set of Tests

In the second set of trials, the focus is on detecting Defect 2. The defect signal is expected to be
received between 4–4.5 m. Since the baseline of the defect signal is highly polluted by the coherent
flexural noise, higher detection limits are required for detecting the defects, even when the CSA size of
the defect is increased. This aim of these tests is to observe the effect of spatial noise on the algorithm
and demonstrate its capability in detecting defects in such scenarios. Since excitation frequencies
of 37 kHz, 34 kHz, and 30 kHz have the lowest spatial noise in the defect region, respectively, it is
expected that the algorithms perform better using the respective mentioned frequencies.
63
Appl. Sci. 2018, 8, 2378
5.3.1. Method One

Figure 17 shows the limits required from method one to detect Defect 2. The only excitation
frequency that allowed the detection of defects without any outliers was 37 kHz, while 34 kHz resulted
in the detection of defects with at least three outliers in both cases, and in 30 kHz, the defect could
not be detected. With 37 kHz as the excitation frequency, it can be seen that when the defect energy
decreases, the threshold is increased to 0.009. As this value is close to the maximum allowable threshold
(0.01), no smaller defects could be detected even using this frequency. This is unlike the first set of
tests, where it was easier to detect the defect using lower frequencies, since Defect 2 is located in a
region with higher energies of flexural noises. Consider the case of the 34-kHz frequency, as reported
in Figure 11: the maximum limit of detection of Defect 1 without any outliers is approximately 0.013.
However, in the case of Defect 2, the minimum level is 0.019 for the 5% CSA defect, which is higher
than the previously reported limit. This outcome indicates the detection of outliers in the test. The
same can also be said regarding 37 kHz, where previously, the maximum limit was approximately
0.012, and now, the minimum required for the detection of a 4% CSA defect is 0.01, which is below the
level where outliers will be detected. It can be concluded that method one is highly affected by the
existing energy of flexurals in the same iteration.

0.03
0.025 0.025
Amplitude (arbt.)
0.02 0.019
0.015 0.01
0.01 0.01 0.0095

0.007
0.005 0.009
0.004
0
30kHz-5% 34kHz-5% 37kHz-5% 30kHz-4% 34kHz-4% 37kHz-4%
Figure 17. Minimum, average, and maximum values of method one of the second test cases.
5.3.2. Method Two

The results of method two are illustrated in Figure 18. Unlike the first set of tests, where 34 kHz
was the optimum frequency of excitation, in these tests, 37 kHz achieves better results. It is not possible
to detect Defect 2 without any outliers using the 30-kHz frequency, since the required minimum
threshold level, 0.125, is higher than the maximums that are reported in Figure 13. Nonetheless, using
the two other frequencies, Defect 2 can be detected. From the margins, it is evident that 37 kHz achieves
the highest safe zone margin required for detection, where both the minimum and maximum thresholds
that are required are respectively lower and higher than the ones required for other frequencies.
Furthermore, when the defect size decreases, the limits are not affected significantly, which suggests
that when using 37 kHz, smaller defect sizes than 4% CSA can be detected. In fact, in this region,
defects with CSAs as small as 2% can only be detected using this frequency using both method two
and method three.
Compared to method one, although the limits have been increased in the case of Defect 2, it can
be seen that it is not affected as much, since method two is capable of detecting defects using 34 kHz
without any outliers and 30 kHz with outliers. Nonetheless, this method also performs best in the
absence of flexurals in the same location as the defect signal.
64
Appl. Sci. 2018, 8, 2378

0.200 0.250
0.180 0.174 0.176
0.173 0.213
Amplitude (arbt.)
0.209 0.211
0.160 0.157 0.200
0.149 0.143 0.186
0.140 0.176
0.128 0.167
0.120 0.125 0.150 0.147
0.107 0.139
0.100 0.098 0.122 0.117
0.092 0.110
0.080 0.100
0.060 0.064
0.068
0.040 0.050
0.020 0.020 0.020
0.011 0.011
0.000 0.000
Figure 18. Limits achieved using method two on the second set of tests.
5.3.3. Method Three

Figure 19 shows the results achieved using method three. These are the same as before in the tests
where the average detection values are in between the minimum and maximum thresholds; namely,
the tests using 37 kHz as the excitation frequency are the tests where the defects can be automatically
detected without any outliers when using this method. Same as the other methods, using 30 kHz for
detection of Defect 2 is only possible when outliers are also reported. For the case of 34 kHz, when
using the average value, it is not possible to detect Defect 2 without any outliers. However, it can be
seen when using the second version of the thresholding value that the minimum detection value is
significantly closer to the average value, which clearly illustrates the advantage of this version.

0.9 0.9
0.85 0.864 0.85 0.864
Amplitude (arbt.)
0.833 0.831
0.8 0.8
0.75 0.691 0.75
0.7 0.691 0.7
0.683 0.683 0.674 0.678
0.65 0.65 0.65 0.664
0.576 0.637
0.5871 0.619
0.6 0.6
0.5933 0.576 0.5863 0.5709 0.5949 0.5933 0.5871 0.572 0.5863
0.5709
0.55 0.5949 0.577 0.572 0.55 0.541 0.539 0.599
0.563 0.519
0.5 0.5
Figure 19. Limits achieved using method three on the second set of tests.
5.3.4. Comparison
Figure 20 shows the number of increments observed in the limits when V2 is used. It is the same
as in the first test case: the minimums tend to increase less than the maximums. In tests using 30 kHz,
since outliers are detected, the same amount of increment is reported.
The generated safe zone margins for the second set of tests are demonstrated in Figure 21. As
expected, V2 always provides a greater safe zone margin. This is especially important in the case of
method three, since generally, the minimums that are required for detection are well above the average
value, while the maximums are close. The benefit of using V2 is clearly illustrated in Figure 19, where
upon testing with the 34-kHz frequency, the maximum detection value of V2 in both tests is much
closer to the calculated average value.
65
Appl. Sci. 2018, 8, 2378
Increment of Minimum Incerement of Maximum
0.044
0.037
0.037
0.037
0.036
Amplitude (arbt.)
0.038
0.037
0.037
0.036
0.029
0.031
0.027
0.019
0.019
0.017
0.017
0.014
0.014
0.005
0.004
0.002
0.000
0.000
0
30KHZ -5% 34KHZ -5% 37KHZ -5% 30KHZ -4% 34KHZ-4% 37KHZ-4% 30KHZ -5% 34KHZ -5% 37KHZ -5% 30KHZ -4% 34KHZ -4% 37KHZ-4%
(a) Method Two (b) Method Three
Figure 20. The increments achieved using the second version of thresholding in the second set of tests.
0.323
0.312
0.287
0.27
Amplitude (arbt.)
0.198
0.193
0.162
0.156
0.108
0.085
0.065
0.059
0.046
0.039
0.033
0.029
0.006
0.001
0
0
0
0
0
0
0
0
0
0
30KHZ -5% 34KHZ -5% 37KHZ -5% 30KHZ -4% 34KHZ -4% 37KHZ-4%
Figure 21. Bar graphs of the Safe zone margins of all methods in the second set of tests.
The average detection lengths achieved using all of the methods are shown in Figure 22 using the
optimum frequency (37 kHz) for detecting Defect 2. Both methods two and three achieve the same
detection length for 5% CSA; nonetheless, when the defect size decreases, method three tends to have a
higher detection length. In the tests with a 34-kHz excitation frequency, method two achieved a higher
detection length, since the average value is not set mathematically. Using the optimum frequency,
it can be concluded that method three is the best method, since it generated the most detection length
with the highest safe zone margin.
0.128
0.128
0.113
0.113
0.103
Amplitude (arbt.)
0.101
0.101
0.101
0.101
0.1
0.1
0.077
0.077
0.055
0.051
0.051
0.023
0.005
0.004
0.004
0.004
0.003
0.003
0.003
0.003
0.003
0.003
0
30KHZ-5% 34KHZ -5% 37KHZ-5% 30KHZ -4% 34KHZ-4% 37KHZ -4%
Figure 22. Bar graphs of detection length of the defect using the average value of all of the methods in
the second set of tests.
6. Conclusions
In this paper, a novel statistical algorithm was proposed that uses the full capability of the tool-set
array of conventional guided waves inspection devices to identify defect signals that have been
polluted by coherent noise. Laboratory trials confirmed the effectiveness of this algorithm in detecting
defects with low/negative SNRs and its dependence on the frequency.
Three different methods and two different versions were created for generating the results,
in which the thresholding scheme of each was different. In method one, the threshold is a static
66
Appl. Sci. 2018, 8, 2378
value set by the inspector. In method two, the assigned threshold to each iteration is a percentage
taken from the summation of the amplitude of all of the transducers signals with the same as the
time-domain signal generated by the normal propagation routine. In method three, the percentage
value is determined by determining the number of transducers with the same phase over the total
number of transducers, subtracted by an offset. Two versions of method two and method three are also
developed, wherein the second version, the summation of signals with the same sign as the temporal
signal was subtracted from the comparisonValue, before determining the final assigned threshold of
each iteration. The following conclusions can be drawn with regards to the tests done:
1. Method Comparison: Method one tended to produce the worst results, as it solely depended
on a static value set by the user without considering the characteristics of the signals received
in each iteration. Methods two and three can achieve a similar result in most cases; however,
in method three, the threshold value can be set automatically, which provides a great benefit in
one-off inspection scenarios.
2. Version Comparison: Version 2 always produced a wider safe zone margin while achieving
approximately the same defect detection length using the average values.
3. Excitation Frequency: In the first set of tests, by lowering the frequency, a greater safe zone margin
and a lower threshold value is required for method one. However, in the second set of tests,
the exact opposite occurs. Therefore, method one is highly affected by the regional flexural waves,
and high torsional energies are required for detection. However, both of the other methods
have an optimum excitation frequency of 34 kHz for the first set of tests, and 37 kHz for the
second set of tests. This suggests that all of the methods are more affected by the existing spatial
noises caused by the flexural waves in the same iteration rather than the temporal noise of the
neighborhood. The second set of tests also clearly illustrates the advantages of method two and
method three in detecting defects in locations where a higher level of coherent noise exists.
4. Safe Margin Zones: Method three (V2) always achieved the highest safe zone margin, and is
therefore the safest method to use.
5. With regard to these results, method three (V2) is the best method since the threshold can be
determined automatically, and it achieves the highest safe zone margin. The tests prove that this
method can detect defects using both 34 kHz and 37 kHz, depending on the existing coherent
noise in the defect location. Furthermore, the algorithm is sensitive to the existing coherent noises
in the baseline signal rather than the temporal noise in the tests.
This newly developed algorithm can be used to help the inspectors as an additional tool, allowing
them an easier interpretation of the results. This is the main reason why all of the developed methods
are mentioned. It also enables the possibility of developing narrowband transducers, as opposed to
the currently used wideband transducers, which have a more focused transfer function and stronger
excitation power.
In this paper, we have defined a new pathway for solving the problem of defect identification
in the UGW inspection of pipelines. In this pathway, we use the information of both space and time
to automatically detect defect signals rather than trying to minimize the coherent noise by spatial
averaging techniques and interpreting the results based on their properties in a time domain. Thus,
the main message of this paper was to illustrate the effectiveness of such approaches, which are feasibly
applicable to the current systems of guided-wave testing.
Author Contributions: Conceptualization, H.N.M.; Data curation, H.N.M.; Formal analysis, H.N.M.; Funding
acquisition, H.N.M.; Investigation, H.N.M.; Methodology, H.N.M.; Project administration, H.N.M.; Resources,
H.N.M.; Software, H.N.M.; Supervision, H.N.M., K.Y. and A.K.N.; Validation, H.N.M.; Visualization, H.N.M., K.Y.
and A.K.N.; Writing—original draft, H.N.M.; Writing—review & editing, H.N.M. and A.K.N.
Funding: This research study was made possible by the sponsorship and support of Brunel University London
and National Structural Integrity Research Centre (NSIRC).
Conflicts of Interest: The authors declare no conflicts of interest.
67
Appl. Sci. 2018, 8, 2378
References
1. Ostachowicz, W.M. Guided Waves in Structures for SHM: The Time-Domain Spectral Element Method; Wiley:
Hoboken, NJ, USA, 2012.
2. Lowe, M.J.S.; Cawley, P. Long Range Guided Wave Inspection Usage—Current Commercial Capabilities and
Research Directions; Imperial College: London, UK, 2016; Available online: http://www3.imperial.ac.uk/pls/
portallive/docs/1/55745699.PDF (accessed on 19 October 2018).
3. Lowe, M.J.S.; Alleyne, D.N.; Cawley, P. Defect detection in pipes using guided waves. Ultrasonics 1998, 36,
4. Thornicroft, K. Ultrasonic Guided Wave Testing of Pipelines using a Broadband Excitation; Brunel University
London: London, UK, 2015.
5. Pavlakovic, B. Signal Processing Arrangment. U.S. Patent 2006/0203086 A1. 14 September 2006.
6. ASTM Standard E2775–16. Standard Practice for Guided Wave Testing of above Ground Steel Pipeword Using
Piezoelectric Effect Transduce; ASTM International: West Conshohocken, PA, USA, 2017.
8. Zeng, L.; Lin, J.; Lei, Y.; Xie, H. Waveform design for high-resolution damage detection using lamb waves.
9. Yücel, M.K.; Fateri, S.; Legg, M.; Wilkinson, A.; Kappatos, V.; Selcuk, C.; Gan, T.H. Coded Waveform
Excitation for High-Resolution Ultrasonic Guided Wave Response. IEEE Trans. Ind. Inform. 2016, 12, 257–266.
[CrossRef]
10. Yücel, M.K.; Fateri, S.; Legg, M.; Wilkinson, A.; Kappatos, V.; Selcuk, C.; Gan, T.H. Pulse-compression
based iterative time-of-flight extraction of dispersed Ultrasonic Guided Waves. In Proceedings of the 2015
IEEE 13th International Conference on Industrial Informatics (INDIN), Cambridge, UK, 22–24 July 2015;
pp. 809–815.
11. Lin, J.; Hua, J.; Zeng, L.; Luo, Z. Excitation Waveform Design for Lamb Wave Pulse Compression. IEEE Trans.
12. Malo, S.; Fateri, S.; Livadas, M.; Mares, C.; Gan, T. Wave Mode Discrimination of Coded Ultrasonic Guided
Waves using Two-Dimensional Compressed Pulse Analysis. IEEE Trans. Ultrason. Ferroelectr. Freq. Control
2017, 64, 1–10. [CrossRef] [PubMed]
13. Song, X.; Ta, D.; Wang, W. A base-sequence-modulated golay code improves the excitation and measurement
of ultrasonic guided waves in long bones. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2012, 59, 2580–2583.
[PubMed]
14. Pedram, S.K.; Haig, A.; Lowe, P.S.; Thornicroft, K.; Gan, L.; Mudge, P. Split-spectrum signal processing for
reduction of the effect of dispersive wave modes in long-range ultrasonic testing. Phys. Procedia 2015, 70,
15. Pedram, S.K.; Fateri, S.; Gan, L.; Haig, A.; Thornicroft, K. Split-spectrum processing technique for SNR
enhancement of ultrasonic guided wave. Ultrasonics 2018, 83, 48–59. [CrossRef] [PubMed]
16. Fateri, S.; Boulgouris, N.V.; Wilkinson, A.; Balachandran, W.; Gan, T.H. Frequency-sweep examination for
wave mode identification in multimodal ultrasonic guided wave signal. IEEE Trans. Ultrason. Ferroelectr.
Freq. Control 2014, 61, 1515–1524. [CrossRef] [PubMed]
17. Ng, C.-T. On the selection of advanced signal processing techniques for guided wave damage identification
using a statistical approach. Eng. Struct. 2014, 67, 50–60. [CrossRef]
18. Clarke, T.; Simonetti, F.; Cawley, P. Guided wave health monitoring of complex structures by sparse array
systems: Influence of temperature changes on performance. J. Sound Vib. 2010, 329, 2306–2322. [CrossRef]
19. Lu, Y.; Michaels, J.E. A methodology for structural health monitoring with diffuse ultrasonic waves in the
presence of temperature variations. Ultrasonics 2005, 43, 717–731. [CrossRef] [PubMed]
20. Chang, Y.; Zi, Y.; Zhao, J.; Yang, Z.; He, W.; Sun, H. An adaptive sparse deconvolution method
for distinguishing the overlapping echoes of ultrasonic guided waves for pipeline crack inspection.
Meas. Sci. Technol. 2017, 28, 035002. [CrossRef]
21. Davies, J.; Cawley, P. The application of synthetic focusing for imaging crack-like defects in pipelines using
guided waves. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2009, 56, 759–770. [CrossRef] [PubMed]
68
Appl. Sci. 2018, 8, 2378
22. Roberts, A.; Pardini, R.; Diaz, A. High frequency guided wave virtual array SAFT. Rev. Quant. Nondestruct.
Eval. 2003, 22, 205–212.
23. Luo, W.; Rose, J.L.; van Velsor, J.K.; Mu, J. Phased-array focusing with longitudinal guided waves in a
viscoelastic coated hollow cylinder. AIP Conf. Proc. 2006, 820, 869–876.
24. Davies, J. Inspection of Pipes Using Low Frequency Focused Guided Waves; Imperial College London: London,
UK, 2008.
25. Davies, J.; Simonetti, F.; Lowe, M.; Cawley, P. Review of Synthetically Focussed Guided Wave Imaging
Techniques with Application to Defect Sizing. In Proceedings of the European Conference on Nondestructive
Testing 2006, Berlin, Germany, 25 September 2006; pp. 1–9.
26. Sicard, R.; Chahbaz, A.; Goyette, J. Guided lamb waves and L-SAFT processing technique for enhanced
detection and imaging of corrosion defects in plates with small depth-to-wavelength ratio. IEEE Trans.
27. Fu, H.; Wu, B.; He, C.; Wang, W. A synthetic digital signal processing method of ultrasonic guided wave
pipeline inspection. In Proceedings of the International Conference on Image Analysis and Signal Processing,
Hubei, China, 21–23 October 2011; pp. 444–446.
28. Meitzler, A.H. Mode Coupling Occurring in the Propagation of Elastic Pulses in Wires. J. Acoust. Soc. Am.
1961, 33, 435–445. [CrossRef]
29. Rose, J.L. Ultrasonic Guided Waves in Solid Media; Cambridge University Press: Cambridge, UK, 2014.
30. Mahal, H.N.; Mudge, P.; Nandi, A.K. Comparison of coded excitations in the presence of variable transducer
transfer functions in ultrasonic guided wave testing of pipelines. In Proceedings of the 9th European
Workshop on Structural Health Monitoring, Manchester, UK, 10–13 July 2018.
31. Lowe, M.J.S.; Alleyne, D.N.; Cawley, P. The Mode Conversion of a Guided Wave by a Part- Circumferential
Notch in a Pipe. J. Appl. Mech. 1998, 65, 649–656. [CrossRef]
32. Nurmalia. Mode Conversion of Torsional Guided Waves for Pipe Inspection: An Electromagnetic Acoustic Transducer
Technique; Osaka University: Osaka, Japan, 2013.
33. Teletestndt, Long Range Guided Wave Testing with Teletest Focus+. 2018. Available online: https://www.
teletestndt.com/ (accessed on 19 October 2018).
69
applied
sciences
Article
Improved Defect Detection Using Adaptive Leaky
NLMS Filter in Guided-Wave Testing of Pipelines
Received: 23 November 2018; Accepted: 8 January 2019; Published: 15 January 2019
Abstract: Ultrasonic guided wave (UGW) testing of pipelines allows long range assessments of pipe
integrity from a single point of inspection. This technology uses a number of arrays of transducers,
linearly placed apart from each other to generate a single axisymmetric wave mode. The general
propagation routine of the device results in a single time domain signal, which is then used by the
inspectors to detect the axisymmetric wave for any defect location. Nonetheless, due to inherited
characteristics of the UGW and non-ideal testing conditions, non-axisymmetric (flexural) waves will
be transmitted and received in the tests. This adds to the complexity of results’ interpretation. In this
paper, we implement an adaptive leaky normalized least mean square (NLMS) filter for reducing
the effect of non-axisymmetric waves and enhancement of axisymmetric waves. In this approach,
no modification in the device hardware is required. This method is validated using the synthesized
signal generated by a finite element model (FEM) and real test data gathered from laboratory trials.
In laboratory trials, six different sizes of defects with cross-sectional area (CSA) material loss of 8%
to 3% (steps of 1%) were tested. To find the optimum frequency, several excitation frequencies in
the region of 30–50 kHz (steps of 2 kHz) were used. Furthermore, two sets of parameters were used
for the adaptive filter wherein the first set of tests the optimum parameters were set to the FEM test
case and, in the second set of tests, the data from the pipe with 4% CSA defect was used. The results
demonstrated the capability of this algorithm for enhancing a defect’s signal-to-noise ratio (SNR).
Keywords: adaptive filtering; leaky normalized mean square; ultrasonic guided waves; pipeline
inspection; SNR enhancement; signal processing
1. Introduction
Ultrasonic guided wave (UGW) allows long-range non-destructive testing (NDT) of pipes from
a single point of inspection. The device utilizes a test-tool consisting of a number of transducer
arrays that are installed on the pipe. One of the main problems of UGWs is that they are inherently
multimodal and dispersive. In general inspection of pipelines, pure axisymmetric modes (torsional and
longitudinal), in their non-dispersive regions, are used to decrease the complexity in the interpretation
of the results. The general propagation routine of the device generates a single time-domain signal
by summing a phase-delayed arrangement of the waveforms received from each transducer [1,2].
The defects are then detected by the inspectors, based on signal-to-noise ratio (SNR), where an anomaly
is reported if the signal envelope is higher than the regional noise level [3].
The main advantage of using a transducer array is the excitation of axisymmetric wave and
cancellation of the flexural waves, which are mainly considered as coherent noise in the inspection.
Nonetheless, due to the imperfect testing condition, perfect cancellation of flexural waves is not
achievable. Various approaches have been established in the literature for improving the SNR in

Appl. Sci. 2019, 9, 294
the presence of coherent noise. One of the methods is pulse-compression [4–7]. In this technique,
a modulated waveform (as opposed to a sine wave) is excited, wherein the reception side results
are correlated with the known sequence. The locations with the highest correlation illustrate the
same possible location of the defects. However, the performance and the design of an optimum
excitation sequence can be dependent on the testing conditions, such as the transfer functions of the
transducer [6]. On the other hand, filtering approaches such as split-spectrum processing have also
been implemented. In 2018, Pedram et al. [8] utilized split-spectrum processing to enhance the SNR of
tests. While significant improvements can be achieved using SSP, the algorithm depends on many
different parameters which can be changed on a case by case basis. Furthermore, wavelet de-noising
has been investigated by Mallet [9]. In this approach, after a defining a mother wavelet, the signal
is decomposed into multiple levels of discrete wavelet transforms where a thresholding technique
is applied to remove low-amplitude signals. The undesirable effect of decreasing the noise with this
method is the possibility of removing the signal of interest, as the non-random noise components
(flexural waves) can have similar amplitudes to those received from axisymmetric waves.
All of the aforementioned techniques can be considered as passive noise cancellation, where a
function must be defined beforehand by the user and results are processed based on that function.
Furthermore, most of the methods in the literature consider the signals achieved after the device’s
propagation routine as the main input signals of their implemented techniques while the device
inherently produces a multidimensional signal. Therefore, in this paper, we propose to use adaptive
filtering which is an active noise cancellation technique. As opposed to passive techniques, active
noise cancellation allows the function to update itself to the noise characteristic of each iteration [10].
In the field of signal processing, filtering is one of the most common processes [11]. When the
characteristics of noise are known, static filters can be designed that remove the non-coherent unwanted
noise of tests (passive noise cancellation). As an example, when the expected bandwidth of the received
signal is between 30 and 60 kHz, all other frequencies can be filtered. However, in most real-life
scenarios, designing an optimum static filter is not possible since: (1) the characteristic of the noise is
not known beforehand and (2) the noise can be iteratively changing. On the other hand, by providing
the right multidimensional signal to adaptive filters, it is possible to remove such noise from the tests.
Adaptive filters have been applied in a variety of industries such as communications radar,
ultrasonic and medical. The basic concept of adaptive noise cancellers was introduced by Widrow et al.
in 1975 [12]. Their work developed the already existing Wiener filter model in order to utilize the residual
signal for adaptive update of the filter weights. Many of the derived algorithms, as well as applications,
were explained in his work, including canceling periodic interference in electrocardiography and
speech signals. Using the same principles, Rajesh et al. [13] proposed an adaptive noise-canceling
method for the removal of background noise where the closest sensor to the abdomen of a pregnant
patient is used as the reference signal, with the maternal signal captured by a sensor in a different
position. By using this configuration, they were able to extract the wanted Fetal signal which was
obscured by the signal received from the maternal heart rate. In 2003 [14], Hernandez implemented
the adaptive linear enhancer (ALE) algorithm to increase the SNR by using the delayed version
of the same signal as the input to the adaptive filter. Ramli et al. [15] reviewed different adaption
algorithms, such as the least mean square (LMS), normalized-LMS (NLMS), leaky-LMS and recursive
least square (RLS), where it was reported all of these approaches are capable of enhancing the SNR.
In communications and speech analysis, one of the common uses of an adaptive noise canceller (ANC)
is for echo cancellation [16,17], which can be generated by hybrid transformers while transmitting
data, or from the multipath mitigation of the signals to the microphone. In radar technology [18,19],
adaptive filters are widely used for multipath cancellation, clutter rejection, and spatial noise reduction
by using adaptive beamforming. In ultrasonic testing, Zhu et al. [20] used adaptive filters in order to
increase the SNR in ultrasonic non-destructive testing (NDT) of highly scattered materials. In 2010,
Monroe et al. [21], tested this configuration for clutter rejection and detection of small targets in
ultrasonic backscattered signals. Both researchers reported that adaptive filtering is a robust method
72
Appl. Sci. 2019, 9, 294
of filtering the random ultrasonic grain noise created by the structure. It was reported that if the
ultrasonic beam is moved less than a beam width away from the position, the echo from a structural
feature is closely related while the grain noise varies. The results were evaluated experimentally using
NLMS where an improvement of 5–10 dB was observed with fixed parameters.
In guided-wave testing, the noise is time varying. While the excited wave mode is non-dispersive
and axisymmetric, the flexural waves change in time. Therefore, when observing the signal from different
points of the pipe the axisymmetric waves will have similar characteristics, while the flexural waves will
be variable. In this paper, instead of removing the noise from the single time-domain signal received from
the propagation routine, a novel method is proposed to utilize adaptive filtering using the signals received
from separate rings to remove the flexural wave modes in the tests. The leaky NLMS method is used for
updating the filter weights in each iteration. The NLMS method reduces the amplification of gradient
noise and also provides a faster rate of convergence than the LMS algorithm [10,22]; therefore, it is more
suited for guided-wave applications with time-varying noise. Furthermore, the introduction of a leakage
factor in the update function for the filter weights limits the divergence of the filter weights. Leakage
methods were initially introduced to decrease performance degradation caused by limited-precision
hardware [23]. The process is equivalent to the introduction of a white noise sequence to the input
signal [22]. In the context of this paper, the main reason for applying a leakage factor was to allow faster
adaption of the filter weights to the existing noise of each iteration. The performance analysis of leaky
NLMS algorithms have been previously done by Sayed et al. [24] and the stability of the algorithm was
investigated by Bismor et al. [25]. In mathematical formulation, introducing the leakage factor would
only reduce the effect of past filter weights. Therefore, if the same principles of the NLMS algorithm are
used, the algorithm will remain stable and bounded. This methodology was initially developed using
synthesized data generated by finite element modelling (FEM) of the test system. It was then verified
through experimental trials on real pipe data with a defect of 3–8% cross-sectional area (CSA) size, and
enhancement of the SNR in comparison to the device’s general routine is demonstrated.
In Section 2, the background of guided waves in pipelines and the nature of noise is explained.
Section 3 includes the description of the device’s general propagation routine and the proposed
adaptive filtering strategy. In the results section, for better illustration of the adaptive filtering process,
each stage of the process applied to the FEM signal is shown, followed by the results achieved from the
experimental tests. It is demonstrated that although the algorithm cannot increase the SNR for small
defects, but it will also not decrease it in compared to general propagation routine.
2. Background Information
Guided waves are multimodal [26]. Depending on the excitation frequency, multiple wave modes
can be generated at the point of excitation. These modes are categorized based on their displacement
patterns (mode shapes) within the structure. The main categories of guided waves in pipes are
axisymmetric torsional and longitudinal waves and non-axisymmetric flexural waves. The popular
nomenclature used for them is in the format of X(n,m), where X can be replaced by letters L for
longitudinal, T for torsional, and F for flexural waves; n shows the harmonic variations of displacement
and stress around the circumference; and m represents the order of existence of the wave mode [27].
In an ideal scenario, a single pure family of axisymmetric waves in their non-dispersive region should
be excited for general inspection. Nonetheless, this is not possible due to the variations of excitation
and the reception transfer function of transducers in real-life scenarios. Furthermore, the interaction of
wave modes with the pipe also leads to the generation of different wave modes, which is commonly
known as a wave mode conversion phenomenon [26,28,29].
Figure 1 shows the spatial arrival of the signals from different segments of the pipe. These signals
are created using the developed model in [30], where a three ring excitation system is used. Each ring
contains 32 source points with variable transfer functions. In the figures, blue and red lines show the
received amplitude from the transducers of two separate rings while the black dotted line serves as
the reference. Any point outside the circle illustrates positive amplitude and the points inside show
73
Appl. Sci. 2019, 9, 294
the negative amplitude of the corresponding source point. The following points can be concluded by
comparing the illustrated polar plots:
• The energy of waves received from axisymmetric features (e.g., pipe end) are overwhelmed by
the T(0,1) with little variance caused by the existing coherent noise of flexural waves.
• Defect signals have a mixture of both flexural and torsional wave modes; nonetheless, if the
cross-sectional area (CSA) of the defect is wide, the T(0,1) will have a greater effect than the
flexurals. This is the main reason why adding all the transducers tends to increase the overall
energy of T(0,1) while reducing the effect of flexurals.
• Noise regions are mostly consistent with flexurals with almost no axisymmetric wave
being detected.
(a) Pipe End (b) Defect (c) Flexurals caused by defect
Figure 1. Example of received FEM signals from the array from different sections of a pipe [31].
Furthermore, guided waves exhibit dispersive propagation. Dispersion causes the energy of
a signal to spread out in space and time as it propagates [32]. Figure 2a shows an example of a
dispersion curve calculated from an 8-inch schedule 40 steel pipe using RAPID software [33]. Wilcox
et al. introduced a method to use dispersion curves in order to both simulate [34] and remove [32] the
effect of dispersion. By applying the developed equation by Wilcox et al. and the shown dispersion
curves in Figure 2a, a simulated propagation of F(4,2) wave mode is created and shown in Figure 2b
where the black and red lines show the same wave mode after 1 m and 2 m of propagation, respectively.
As can be seen, for a dispersive waveform the temporal resolution and the energy of wave packet is
decreased. Therefore, in general inspection, non-dispersive regions of wave modes are excited to allow
ease of inspection. In this research, T(0,1) is used as the excitation wave mode since it is non-dispersive
across its whole frequency range [35].
7 ) ) )

) ) )

6SHHGPV

)UHTXHQF\N+]
(a) (b)
Figure 2. (a) Example dispersion curve of T(0,1) wave mode in an 8-inch schedule 40 steel pipe. (b)
The effect of dispersion on a simulated flexural wave, two propagation distances [31].
74
Appl. Sci. 2019, 9, 294
Three main signals are expected to be received in guided-wave tests which are shown in Figure 3.
The first signal is the superposition of T(0,1) and flexurals where it is the true defect location.
The second is the flexurals created due to wave mode conversion [36]; as these signals exhibit the same
characteristics of excitation waveform, they can lead to the detection of false alarms. The third category
is the non-coherent white noise, which is not correlated to the excitation sequence. This category is
the main one which can be removed using generalized filtering techniques, as their characteristic is
completely different to the excitation sequence.
White Noise (Non-Coherent)
Superposition of signal
Coherent Noises
and multiple modes
Figure 3. Categories of the signals in the guided-wave inspection [31].
3. Methodology
As explained previously, guided waves are multimodal. To allow inspection of pipelines using
guided waves, a single wave mode must be excited in only one direction to allow detection of the
anomaly’s location. Excitation and detection of flexural waves are difficult and require complex
processing algorithms due to their circumferential variation. Therefore, the practice in the general
inspection is to excite an axisymmetric wave. For doing so, the toolset of guided-wave testing devices
consists of arrays of linearly placed transducers (rings) across the circumference of the pipe. This leads
to excitation of a quasi-axisymmetric wave where, in ideal scenarios, the corresponding flexurals of the
excitation wave mode will cancel each other out [1,37,38]. If only one array is used, a bidirectional wave
mode is generated since there is no directional control in the normal piezo-electric transducer. The most
common method of achieving a unidirectional transmission is to use multiple arrays of transducers.
By considering the dispersion curves of the generated wave mode and the placement spacing between
the rings, waves can be manipulated in such a way to enforce the forward going propagation and
suppress the backward leakage. Hence, the forward propagations of each ring are overlapped while
the signals received from backward propagation direction are inversely overlapped [1,2,6,39].
On the reception side, signals from the backward direction will still be received due to (1) the
impure/non-ideal excitation and (2) the forward propagation signals travel paths after the toolset.
Once the device is in monitoring mode, no excitation takes place, meaning if an echo is received from
the device, it will continue to travel past the test tool; if any anomalies exist in that direction, a new
echo will be generated which will travel in the same way as the forward testing direction. At this point
two scenarios will occur:
1. In the first scenario, the signal from a feature in the backward test direction is detected by the
rings. The energy of this signal can be reduced by using the same phase-delaying algorithm in
the excitation sequence.
75
Appl. Sci. 2019, 9, 294
2. The second scenario is when the same echo is past the tool. This signal will have the same
characteristics as forward propagation and thus cannot be canceled. These are typically known
as mirror signals.
The phase-delaying algorithm is once again applied in the general routine of reception to reduce
the effect of signals generated from the first scenario. Afterward, the average of all transducer signals is
generated, which will reduce the energy of flexural waves. The signal of the rings can also be strongly
different which either suggests the existence of no defect or a small defect which is not generating
a strong-enough signal to be detected. In the normal approach, as simple summation is used, this
will lead to the generation of a residual signal which might be mistaken as a defect in the inspection.
The flowchart for this general reception routine, which is the common practice of industry, is shown in
Figure 4.
First set of signals
Second set of signals
Figure 4. Flowchart of general propagation routine currently used in guided-wave devices.
Adaptive Filtering
The main challenge in the inspection of guided waves is to decrease the energy of coherent noise in
the test since non-coherent noise is generally removed by bandpass filtering and averaging. Dispersion
characteristics of guided wave modes vary with regards to the structural features, frequency, and wave
mode. However, the non-dispersive torsional mode is typically constant with an approximate speed of
3260 m/s. By shifting the rings to overlay the T(0,1) signals together, the wave modes which travel
at different speeds will be less correlated with those that travel at constant speed. Furthermore, as
shown in Figure 1, the flexural waves are highly variable in the space while the torsional mode has a
higher correlation.
In this case, the existing torsional mode is highly correlated while the noise and flexural modes
will be less correlated in time. Therefore, instead of the summation of the signals, adaptive filters can
be utilized to enhance the signal resolution. The output signal of an finite impulse response (FIR) filter
in time-domain can be achieved by the convolution operation using [10,40–42]:

p
y(n) = wk ( n ) × ( n − k ) (1)
k=0
where p is the filter length, n is the iteration number, wk (n) is the filter coefficients with the index of
k, and x(n) is the input signal. In the following equations, both x(n) and w(n) are vectors of size p
(represented by bold characters). The resultant error signal would be:
e(n) = d(n)−y(n) (2)
where d(n) represents the required reference (desired) signal; hence, in order to use this method, at
least two signals are required. In adaptive filtering, the goal is to minimize the error signal by setting
the filter weight. Therefore, in each iteration the filter weights should be updated:
w(n + 1) = w(n) + Δw(n) (3)
76
Appl. Sci. 2019, 9, 294
where Δw(n) is the correction step applied to previous filter coefficients to update them for the next
iteration; the key component of the adaptive filter is to define Δw(n) in a way that it decreases the
mean square error over time. One of the robust methods is the normalized least mean square [20,41],
which tries to minimize the least mean square error by setting the coefficients according to a variable
step size. The formula is as follows:
x(n)
w(n + 1) = w(n) + β 2 e(n) (4)
+ x(n)
where β is a normalized step size with 0 < β < 2, is some small positive number to bypass the problem
2
when x(n) becomes too small. The noise in guided waves is highly variable due to the existence of
time-varying multiple modes. To make sure the filter is not over-fitted to the characteristics of previous
iterations, a leakage factor, α, is added to reduce the effect of previous iterations [42]:
x(n)
w(n + 1) = αw(n) + β 2 e(n) (5)
+ x(n)
The flow chart of this algorithm is shown in Figure 5. In this setup, instead of adding the signals
received from the rings of transducers, they are selected as the desired response d(n) and the input
signal x(n). Every other signal is updated based on the characteristics of each iteration and y(n) is the
output of the filter. The selected reference and input signals are the two sets generated in the general
propagation routine, as marked in Figure 4. Applying this phase delay and summation of arrays
within each ring, two sets of signals are created where both have the following components:
• S(t) Signals of interest/anomalies: Highly correlated and in phase

• N(t) Flexural and other noise: That are not correlated to S(t). Furthermore, noise sources of two
signals are correlated to each other.
Update algorithm for each Filter-Weight
‫ܖܗܑܜܘ܉܌܉‬
‫ܟ‬ሺ࢔ െ ૚ሻ
‫ܟ‬ ࢻ
ሺሻ
Ⱦ ሺሻ
Ԗ൅ȁሺሻȁʹ
‫ܟ‬ሺ‫ܖ‬ሻ
Figure 5. Adaptive linear prediction algorithm for noise cancellation where the red marked parameters
are fixed by the user.
This satisfies the need for a reference signal in using adaptive filtering. To cancel the backward
leakage, this algorithm is done with the switching of the input and reference signals, where the final
result is the summation of the two generated outputs of the adaptive filters.
4. Test Setup
The algorithm was initially developed on a synthesized data generated by FEM using ABAQUS
software [30,43]. It was then validated in laboratory trials on real pipes using a Teletest Device [44].
77
Appl. Sci. 2019, 9, 294
4.1. Finite Element Model

The simulated pipe is based on the characteristics of an 8-inch schedule 40 steel pipe with a length
of 6 m, outer diameter of 219.1 mm, thickness of 8.18 mm, Young’s Modulus of 210 GPa, density of
7850 kg/m3 and Poisson’s Ratio of 0.3. The excitation frequency in the test was 30 kHz. In order to
capture the smallest wavelength in the operating frequency, at least 8 elements are required in the axial
direction [45,46]. Therefore, the global seed size was set as 2.5 mm and the time increment was set
as 3.11 × 10−7 s to provide a satisfactory accuracy in generation and reception of all the wave modes
within the frequency band. The modeled test tool included 3 rings of 32 linearly spaced source points,
located 1 m away from the back end of the pipe. Each source point of the rings is numbered clockwise
starting from 90◦ . The schematic of this setup is illustrated in Figure 6.
Test Tool located at 1m
Front End

Back End (BE)
Defect of 3% cross-sectional area size at 4m
Figure 6. Schematic of the FEM with test tool and defect located at 1 m and 4 m, respectively, from the
back end [30].
This type of FEM synthesis of guided wave signals in pipes was developed in previous studies [37,
46–50]. The generated models have also been validated by confirming the accuracy of the received
signals from both the device [50], and 3D laser vibrometer [46,47]. Nonetheless, this type of FEM
generally produces a limited amount of noise due to the perfect condition of the model. The main
aim of this paper and the developed model is to investigate the performance of the algorithm in the
presence of coherent noise. Therefore, in order to generate flexural noise in the test, different transfer
functions were applied to the excitation sequence of each source point. Since the minimum satisfactory
number of elements is used, the transfer functions will not affect the integrity of the model and all the
existing wave modes in the test will be captured. Apart from these transfer functions, all other limits
and parameters were in line with the verified models in the previous studies.
In 2013, Engineer [51] extracted the broadband frequency response of transducers with regards to
various coupling forces using Teletest piezo-electric transducers. The reported frequency response
78
Appl. Sci. 2019, 9, 294
was used as a reference and variability in terms of amplitude and phase was introduced to generate a
variable transfer function for each transducer. An example is shown in Figure 7.
Ideal Ideal Ring1-Tx1

Ring1-Tx1 Ring1-Tx2
Ring1-Tx2 Ring1-Tx4Ring1-Tx4 Ideal Ideal Maximum
Maximum Average
Average Minimum Minimum
1.1 1.1
Normalized Amplitude (arbt.)
Normalized Amplitude (arbt.)

1
11.11 1
0.9 Variable peak 0.9
0.8 resonances Variance
0.7
1 0.8
between
0.7
0.6 0.6
0.5 0.9
0.5
0.4
0.4
0.3 0.8 0.3
0.2
0.2
20 22 24 26 28 30 32 34 36 38 40 20 22 24 26 28 30 32 34 36 38 40
Frequency (kHz) 0.7 x 1000 Frequency (kHz) x 1000
(a) (b)
06
Figure 7. Example of designed transfer functions wherein (a) three different source points are compared
and (b) the maximum, average, and minimum amplitudes across all of the applied transfer functions
used in an individual test are shown.
The transfer functions are multiplied by the ideal excitation’s frequency response, which is a 10
cycle 30 kHz Hann-windowed sine wave, and then inverted back into the time domain. The generated
time-domain signals are used as input files for each of the source points. Additionally, the signals of
the second array are phase delayed and inverted with regards to the 30 mm ring spacing cancellation
algorithm [1,6]. The overall system works in the pulse-echo mode and the signals were received from
the same ideal point sources. The signal generation, as well as the post-processing, were done using
MATLAB-R2016a [52]. In order to increase the coherent noise of testing conditions, instead of receiving
the signals from all source points, only source point numbers 1, 7, 14 and 26 of each ring have been
used for reception (marked in red in Figure 8).
11.25°

Figure 8. The used reception points of each ring in the FEM.
The signal routes of the FEM test are shown in Figure 9. As can be seen, the first directly received
signal from the forward test direction is the defect signal which is received at 3 m, followed by the
front-end signal at 5 m. However, as the transfer function and reception points are not ideal, a backward
leakage signal will exist, propagating in the forward direction. This leakage would be received at 1 m
but due to the backward cancellation algorithm, it will not be observable. However, the reflection
of this signal with the defect signal would be received from the forward direction, leading to a false
79
Appl. Sci. 2019, 9, 294
alarm at 4 m. In these tests, since both signals at 3 and 4 m have the same characteristics, they are both
considered defect signals even though the second signal is a false detection.
Forward Test Direction
Figure 9. Torsional mode reception route in the FEM test cases.
4.2. Experimental Trials

In the laboratory trials, a Teletest MK4 device [44] was used in order to inspect an 8-inch schedule
40 steel pipe with a length of 6 m. The test tool included a collar of 3 ring transducers with 30 mm
ring spacing, located 1.5 m away from the back end of the pipe. Each ring included 24 linearly spaced
thickness-shear (d15) piezoelectric transducers. The system provided both excitation and reception of
signals in pulse-echo mode with the operating frequency of 20 to 100 kHz. The device can capture the
transducers’ signals with a sampling frequency of 1 MHz. On the reception side, an internal bandpass
filter was applied to filter any frequencies outside the main bandwidth of the excitation bandwidth.
The device outputs the time-domain signal received from each ring, as well as the processed signal
with the general propagation routine. In this paper, the signals from the rings are extracted in order to
apply the algorithm.
The defects in these trials are located 4.5 m from back end of the pipe (3 m away from the test
tool). The defects were introduced by removing the pipe’s surface material with a saw. The defect
sizes were varying from 3% to 8% CSA with steps of 1% CSA (total of 6 defect sizes). The signal routes
of the laboratory trials are shown in Figure 10. As opposed to the FEM test setup, the Defect-2 signal,
which was generated by reflection of backward leakage signal from the defect, is now overlapped with
the front-end signal and is no longer detectable.
Forward Test Direction
Figure 10. Torsional mode reception route for laboratory trials.
80
Appl. Sci. 2019, 9, 294
5. Results
In order to better understand this process, an example test case is initially shown where each
stage of the algorithm is explained thoroughly. Then the experimental results are demonstrated in the
second part of this section where the SNRs are calculated with regards to defects with different CSA
sizes and excitation frequencies. In both cases the SNRs are calculated based on the following formula:
⎛ ⎞
⎜⎜ ⎟
⎜⎜ N 1 Signaln ⎟⎟⎟
1 N
SNR = 20 × log 10⎜⎜⎜⎜ ⎟⎟
⎟⎟ (6)
⎜⎝ 1 M ⎟
M Noise ⎠
1 m
where Signal is considered the window where the energy of the signal is received (between 3.0 m and
3.5 m depending on the frequency) and everything else is considered noise. In both tests, excitation
sequences are 10-cycle Hann-windowed sine waves.
5.1. Example Test Case

This technique is initially developed and validated on synthesized data generated by FEM.
The used excitation sequence for generating this data set is 30 kHz and variable transfer functions were
used for each of the source points. Figure 11 shows the two sets of signals achieved after applying
the phase delays of general propagation routine (blue and red). Four regions are of main interest in
this figure, which are (a) backward leakage, (b) flexural noise, (c) Defect-1 and (d) Defect-2. Since
both Defect-1 and Defect-2 signals are received from the forward testing direction, they have similar
characteristics and thus both should be enhanced by this algorithm. As can be seen, noise region (b) is
significantly less correlated than the defect signals (c and d). Furthermore, although the amplitudes
of two sets of signals received in the defect regions are not the same, their phases and polarity are in
perfect correlation. Therefore, while in the propagation routine their characteristic might be similar
but, in this space, the defect and flexural signals are distinguishable.
(a) Backward
Leakage
(d) Defect 2
(c) Defect 1
(b) Flexural Noise
Figure 11. FEM signals (30 kHz) where two defect regions and the flexural noises are marked where
blue and red lines show the first and second sets of the signals respectively.
81
Appl. Sci. 2019, 9, 294
5.2. Cancellation of Backward Leakage

Figure 11a shows the two sets of signals received from the backward testing direction. As can
be seen, these signals are inversely overlapped with little differences in amplitude. This is due to the
general propagation routine where the received signals from rings are delayed and inverted depending
on the ring spacing.
After applying the filtering, the phase of the outcome does not change, therefore by switching the
inputs to the adaptive filter, two sets of results will be generated in which the signals from the forward
direction are overlapped and the ones from the backward direction are inversely overlapped. This is
illustrated in Figure 12, where blue and red signals show the two sets of results achieved from adaptive
filtering by switching the inputs. This operation not only enhances the defect signals received from the
forward direction (Figure 12c) but it also decreases both the backward cancellation (Figure 12a) and
flexural noise regions (Figure 12b) significantly.
(a) (b) (c)
Figure 12. Two sets of results after swapping the inputs to adaptive filter where (a) shows the signals
received from the backward direction, (b) shows the flexural noise, and (c) shows the defect signal.
5.3. Adaption of Filter Weights

The used filter order in the tests is five, which means only five past samples are required to process
the results (Appendix A). Since the number of available samples is limited, observation of the filter’s
frequency response is not possible. Therefore, in Figure 13b–f, the changes of the first filter value
are shown while Figure 13a shows their corresponding results in the time-domain. The red signals
illustrate the results of the leaky-NLMS while the black dotted lines represent the normal NLMS
algorithms. While normal NLMS depends more on past values and is not fast to adapt to the current
iteration, leaky NLMS makes this adaption faster by reducing the filter weights’ dependence on their
past values. After each adaption to a big sequence such as backward leakage or defect, the magnitude
of the filter weight tends to move toward zero; in other words, the weights forget the characteristics of
past iterations significantly faster, which allows them a quicker adaption to the characteristic of the
current iteration. Furthermore, as illustrated by blue dotted signals, in the region of defects, a high
peak is generated in the magnitude of each filter order set by the leaky-NLMS method. Although
other techniques and updated algorithms exist, this fast adaption algorithm, where the filter can forget
the characteristics of the previous wave, is the main reason why leaky-NLMS is more effective in the
guided-wave inspection of pipelines.
82
Appl. Sci. 2019, 9, 294
(a) (b)
(c) (d)
(e) (f)
Figure 13. Comparison of the result achieved from NLMS (black) vs leaky LMS (orange) where
(a) shows the time-domain results and (b–f) show the normalized magnitudes of filter order 1 to
5, respectively.
5.4. Results Comparison

The performance of the algorithm was assessed in experimental trials by detecting defects of
CSA sizes between (and including) 3% and 8%. It is expected that the algorithm would be more
effective at lower frequencies since they are more dispersive; furthermore, the excitation power of
lower frequencies is higher than those of higher frequencies, both due to the duration of the signals as
well as the systems transfer function. Therefore, frequencies between 30 to 50 kHz with a step of 2 kHz
were tested. The SNR achieved from the general propagation routine is demonstrated in Appendix B.
The peak SNR is always achieved using 40 kHz excitation frequency where the SNR decreases by
moving toward either side of that frequency. Furthermore, the SNRs achieved from each frequency are
reduced almost linearly when the defect size is decreased; thus, the noise energy is almost constant
while the defect energy is decreasing.
83
Appl. Sci. 2019, 9, 294
5.4.1. Parameter Selection

The results were generated from two sets of parameters (given in Appendix C Table A3):
• Model Parameters: The optimum parameters are achieved by using a brute force search algorithm
to find the parameters which give the maximum SNR in the FEM test. The main goal of this
test was to assess the response of the algorithm by fixing the parameters to an optimum solution
created by the FEM.
• Experimental Parameters: The brute force search algorithm was performed on the 4% CSA sample
to find parameters that result in most enhancements of SNR for each frequency. The goal was
to find the best testing frequency where the least variations in the enhancements are observed
and assess whether fixed parameters can enhance the SNR of defects with lower CSA size (3%
test case).
5.4.2. Model-Parameters
The results achieved from model parameters are shown in Figure 14 where (a) shows the achieved
SNR values and (b) shows the amount of enhancement. The following conclusions can be extracted
from these tests:
• In defects above 5% CSA, the algorithm enhances the results using all frequencies. For defects
with CSA size of lower than 4%, the algorithm can enhance the SNR of most frequencies except
for 30, 32, 34, and 38 kHz.
• Except from the case of 3% CSA error, both maximum SNR and maximum enhancements are
achieved using 34 and 38 kHz in the tests. In the 3% CSA test, the maximum enhancement is for
30 kHz; nonetheless, the final SNRs of 34 and 38 kHz are greater.
• The variations in the case of 30 kHz are less in comparison to all other cases, as shown in Figure 14b.
Figure 14. Results achieved using fixed model parameters for all frequencies where (a) shows the achieved
SNR and (b) shows the amount of improvement in comparison to the general propagation routine.
5.4.3. Experimental Parameters

The results of experimental parameters are illustrated in Figure 15. The following conclusions can
be drawn from these tests:
• For defects with higher CSA error size (above 5%), lower frequencies tend to achieve higher
SNR after processing where the defect signal was enhanced by at least a factor of 2. The greatest
84
Appl. Sci. 2019, 9, 294
enhancement is achieved from 34 kHz and the maximum SNR is achieved from 36 kHz for
these signals.
• Although higher frequencies will result in lower SNR, their enhancements are approximately the
same for different sizes of the defect. Nonetheless, their final SNR is almost always lower than all
other frequencies, especially in the cases of frequencies above 46 kHz.
• For 3% and 4% CSA defects, the greatest enhancement and gain are achieved using the 38 kHz
frequency. Furthermore, as can be seen in Figure 15b, this frequency is affected less with regards
to the CSA loss of the defect.
Figure 15. Results achieved using experimental parameters, which are variable for each frequency,
where (a) shows the achieved SNR and (b) shows the amount of improvement compared to the general
propagation routine.
5.4.4. Discussion
Figure 16 shows the difference of enhancements between the results of model parameters and
experimental parameters. These results show that although temporal SNR of the signals is important,
the spatial characteristic of the noise also plays an important role in the amount of enhancement. As an
example, while 40 kHz has the highest SNR from the general routine, it does not perform as well as
38 kHz for the algorithm. In most cases of lower CSA defects, the SNR improvements are achieved
not by enhancing the axisymmetric signal strength, but by removing the flexural noise. Except for
excitation frequencies of 30, 34, and 40 kHz in the 3% CSA defect, all other cases were enhanced when
the experimental parameters were used.
Since signals with SNRs higher than 30 dB are more easily detected, it is better to use a frequency
which provides a stable enhancement for both low and higher CSA sizes rather than setting parameters
which can greatly increase the higher CSA defects while decreasing the SNR of smaller defects.
An excitation frequency of 38 kHz provides approximately the same enhancement using both model
parameters and experimental parameters, which shows that this frequency is less affected by the
parameters in comparison to the others. Furthermore, it can also be seen from the results of model
parameters that 38 kHz always achieves the maximum gain even though the parameters are optimized
for 30 kHz excitation frequency in the model.
85
Appl. Sci. 2019, 9, 294
Figure 16. Amount of improvement achieved by experimental parameters as opposed to model parameters.
All frequencies above 40 kHz fail to provide any improvements for defects with CSA sizes smaller
than 4% using model parameters. Nonetheless, even the enhancements achieved with experimental
parameters in those frequencies are much less than all other frequencies, which suggests that this
algorithm tends to perform better at lower frequencies. This is expected since flexurals with lower
frequencies are highly dispersive, which adds to the spatial noise of the tests.
It should be born in mind that, even though in some frequencies the mathematically calculated
SNR might be less than the one achieved from the general propagation routine, the high-order flexurals
are always canceled, which leads to easier interpretation of results by the inspectors. Figure 17 shows
the results of the 30 kHz excitation sequence for defects with different CSAs. As can be seen, highly
variable noises in the region between 2.5 and 3 m are canceled but those caused by low order flexurals
at 2 to 2.5 m are enhanced. While in defects with higher CSA sizes most improvement is achieved due
to the amplification of the defect signal, in lower CSA defects (especially in the case of 3% CSA), the
noise cancellation is the main reason for enhancement. The enhancement of these low order flexurals
is the main reason why the algorithm fails to improve the SNR in some of the testing frequencies.
With regards to the dispersion curves (Figure 2a), low order flexurals become non-dispersive on
frequencies above 30 kHz with their speed moving toward that of T(0,1), 3260 m/s. Since the phase
delays are applied with respect to T(0,1) wave speed, if the flexural waves have the same speed they
will be overlapped in the input and desired responses. This is the main reason why such signals might
not be canceled and, in fact, in some scenarios, they might even be amplified. Furthermore, considering
the 34 kHz filtered signal shown in Figure 18b, the noise envelope is clearly dispersed (2–2.5 m).
Figure 19 also shows the results achieved from 38 kHz excitation frequency. In these figures, the
input and reference signals for the noise and defect regions are shown as blue and red. The following
conclusions can be extracted by observing the difference between a defect and noise signals:
1. If any low order flexurals or noise are overlapping in the input and reference signals, they will
not be removed and might be amplified.
2. The amplification of the signals is also dependent on the amplitude of the input signals to the
algorithm, e.g., the defect signal in case Figure 18 does not observe the same enhancement of that
in case Figure 19 since the blue signal in case Figure 19 is much stronger.
3. Although some flexural signals will remain in the resultant signal, there is an option of further
post-processing to enhance the SNR. Such is the case of the noise region located at 2–2.5 m in
case Figure 18, where it is clearly dispersed at the end of its window, and the case of 1.5–2 m in
86
Appl. Sci. 2019, 9, 294
Figure 19, which results from backward leakage. Therefore, by considering other characteristics
in the signal, the results can be post-processed to distinguish different modes and further increase
the SNR of the test.
(a) 8% CSA (b) 7% CSA
(c) 6% CSA (d) 5% CSA
(e) 4% CSA (f) 3% CSA
Figure 17. Example of filtered signals using 30 kHz excitation and model parameters where defect size
is varying.
87
Appl. Sci. 2019, 9, 294
Defect
Noise Region
(a) General Propagation Routine
(b) Filtered Signal
Figure 18. The result of (a) general propagation routine vs (b) filtered signal with their corresponding
inputs (blue and red lines) of noise (left side) and defect (right side) with the excitation frequency of
34 kHz gathered from an experimental pipe with defect size of 3% CSA.
Defect
Noise Region
(a) General Propagation Routine
(b) Filtered Signal
Figure 19. The result of (a) general propagation routine vs (b) filtered signal with their corresponding
inputs (blue and red lines) of noise (left side) and defect (right side) with the excitation frequency of
38 kHz gathered from an experimental pipe with defect size of 3% CSA.
88
Appl. Sci. 2019, 9, 294
6. Conclusions
In this paper, an adaptive filtering approach is utilized to enhance the SNR of the torsional waves
generated from defects in guided wave inspection of pipelines. For doing so, as opposed to using a
general propagation routine of guided-wave testing devices, the two sets of signals received from the
test tool were used as inputs to the adaptive filter, which resulted in a single output in the time domain.
Inherently, in these two sets of signals, the torsional waves are overlapping while the flexurals are
varying; this was the basis on which the adaptive filter was able to enhance the SNR of torsional waves.
Since flexural waves are variable both in space and time, the used updated algorithm for filter weights
was leaky NLMS, which allowed faster adaption to the highly variable flexural noise and enabled a
faster reset of the filter weights.
The algorithm was initially developed and validated using the synthesized signal generated from
an FEM test case. Afterward, laboratory trail validation took place with two sets of filter parameters:
model parameters, where the parameters were set to find the highest SNR for the FEM test case with
excitation frequency of 30 kHz; and experimental parameters, where the optimum parameters for each
frequency were generated with regards to the real pipe data with 4% CSA defect. In most test cases,
all the frequencies resulted in enhancement of the SNR. Nonetheless, the optimum testing frequency
was found to be 38 kHz since it achieved both the maximum and greatest enhancement of SNR in
tests with defects. Also, this excitation frequency was less affected by the change of parameter as
limited enhancement was observed between the results achieved by using model parameters and
experimental parameters.
Comparing the results from model parameters and experimental parameters, it is evident
that setting the parameters with regards to each individual frequency enhances the achieved SNR.
Nonetheless, optimum selection of parameters is a trade-off between the maximum achieved gain
and the stable enhancement of the SNR of smaller defects. This was assessed using experimental
parameters; when the optimum parameters for 4% CSA defect were used for the case of 3% defect, not
all frequencies could improve the SNR. The model parameters test proved that the parameters can be
chosen using an FEM model, but it will not result in the maximum gain.
The amount of SNR achieved in each test case depends not only on the enhancements of the
torsional wave but also the cancellation of the flexural waves. It must be borne in mind that since the
physical arrangement of the rings is fixed, in some cases, especially for lower order flexural waves that
have closer wave speeds to torsional waves, noise will not be canceled. However, the higher order
flexural noise will be canceled in all the tests and the achieved results can be further processed for
torsional wave mode detection.
Author Contributions: Conceptualization, H.N.M. and A.K.N.; methodology, H.N.M. and A.K.N.; software,
H.N.M.; validation, H.N.M.; formal analysis, H.N.M.; investigation, H.N.M.; resources, H.N.M.; data curation,
H.N.M.; writing—original draft preparation, H.N.M.; writing, reviewing and editing, H.N.M., K.Y., and A.K.N.;
visualization, H.N.M. and A.K.N.; supervision, K.Y. and A.K.N.
Funding: This publication was made possible by the sponsorship and support of Brunel University London and
National Structural Integrity Research Centre (NSIRC).
Appendix A
The filter order was selected based on the FEM data. A brute force search was applied to maximize
the SNR of Defect-1, calculated by Equation (6). Therefore, the best case of each weight is considered.
The generated results are shown in Table A1. The SNR of the unfiltered signal is 5.3 dB. Therefore, this
signal is more difficult to detect as it is almost buried in the noise level. As can be seen, by increasing
the filter order, the SNR is reduced. When the size of filter order grows above 8, the SNRs of filtered
and unfiltered signals are approximately the same. This is because by growing the number of samples,
the differences between a torsional and flexural wave become limited as they are in the same frequency
band. However, in smaller steps, the changes in flexural waves are more time-varying. Therefore, from
89
Appl. Sci. 2019, 9, 294
these results, 8 can be the considered as the safest number for filter order where no change will be
applied to the signal and 2 can be considered as the best filter order for maximizing the SNR.
On the other hand, the best filter order can be dependent on other factors, such as frequency and
the existing flexurals of the iteration. Since the main aim of this paper is to test the workability of
the adaption algorithm to the guided-wave scenario, the filter order is fixed to 5. This number is a
compromise between the safest choice, 8, and the best achieved results, 2, based on FEM with 30 kHz
excitation sequence. Furthermore, a filter order of 5 can be considered as a good compromise, as a
drop of only 1 dB is observed from the maximum achieved SNR, which is 8.31 dB.
Table A1. Maximum SNR achieved for each filter order on FEM test signal with 30 kHz signal.
Filter Order (#) 2 3 4 5 6 7 8 9 10

Maximum SNR
8.31 8.27 7.76 7.32 6.78 6.07 5.33 5.07 5.02
(dB)
Appendix B
The SNR of the defect signal achieved from the general propagation routine of the device is
demonstrated in Table A2 and Figure A1.
Table A2. SNR achieved from the general propagation routine (in dB).
CSA Frequency (kHz)

(%) 30 32 34 36 38 40 42 44 46 48 50
8 14.68 16.81 19.64 21.04 21.73 22.28 21.84 20.47 18.33 15.70 12.63
7 13.55 15.50 18.19 19.69 20.64 21.39 20.97 19.48 17.20 14.53 11.38
6 11.70 13.57 15.89 17.88 18.78 19.33 18.75 17.27 15.06 12.40 9.24
5 9.29 11.09 13.71 15.00 16.05 16.74 16.05 14.40 12.07 9.38 6.17
4 7.13 8.84 11.12 12.15 13.13 13.76 13.23 11.73 9.50 6.84 3.50
3 4.29 5.97 8.15 8.62 9.08 9.51 9.00 7.60 5.43 2.82 −0.51
Figure A1. SNR of tests using general propagation routine.
90
Appl. Sci. 2019, 9, 294
Appendix C
Table A3 shows the parameters used in each test with regards to the frequency.
Table A3. Used parameters in the tests.
Frequency (kHz) Step Size Leakage Compensation

30 (Model) 0.01 0.96 0.01
30 0.01 0.61 0.21
32 0.01 0.46 0.21
34 0.01 0.46 0.26
36 0.01 0.56 0.06
38 0.16 0.81 0.01
40 0.86 0.76 0.01
42 1.41 0.91 0.01
44 1.21 0.96 0.01
46 1.46 0.96 0.01
48 1.46 0.96 0.01
50 1.46 0.96 0.01
References
2. Pavlakovic, B. Signal Processing Arrangment. US Patent 2006/0203086 A1, 14 Septenber 2006.
3. ASTM Standard E2775-16. Standard Practice for Guided Wave Testing of above Ground steel Pipeword Using
Piezoelectric Effect Transducer; ASTM: West Conshohocken, PA, USA, 2017.
[CrossRef]
5. Fateri, S.; Boulgouris, N.V.; Wilkinson, A.; Balachandran, W.; Gan, T.H. Frequency-sweep examination for
wave mode identification in multimodal ultrasonic guided wave signal. IEEE Trans. Ultrason. Ferroelectr.
Freq. Control 2014, 61, 1515–1524. [CrossRef]
6. Thornicroft, K. Ultrasonic Guided Wave Testing of Pipelines Using a Broadband Excitation. Ph.D Thesis,
Brunel University London, London, UK, 2015.
2017, 64, 1092–1101. [CrossRef] [PubMed]
enhancement of ultrasonic guided wave. Ultrasonics 2018, 83, 48–59. [CrossRef]
9. Mallet, R. Signal Processing for Non-Destructive Testing. Ph.D Thesis, Brunel Unviersity, London, UK, 2007.
10. Hayes, M.H. Statistical Digital Signal Processing and Modeling; John Wiley & Sons, Inc.: New York, NY, USA,
1997.
11. Bellanger, M.G. Adaptive Digital Filters, Revised and Expanded, 2nd ed.; Marcel Dekker, Inc.: New York, NY,
USA, 2001.
12. Widrow, B.; Glover, J.R.; McCool, J.M.; Kaunitz, J.; Williams, C.S.; Hearn, R.H.; Zeidler, J.R.; Dong, J.E.;
Goodlin, R.C. Adaptive Noise Cancelling: Principles and Applications. Proc. IEEE 1975, 63, 1692–1716.
[CrossRef]
13. Rajesh, P.; Umamaheswari, K.; Kumar, V.N. A Novel Approach of Fetal ECG Extraction Using Adaptive
Filtering. Int. J. Inf. Sci. Intell. Syst. 2014, 3, 55–70.
14. Hernandez, W. Improving the response of a wheel speed sensor using an adaptive line enhancer. Measurement
2003, 33, 229–240. [CrossRef]
15. Ramli, R.M.; Noor, A.O.A.; Samad, S.A. A Review of Adaptive Line Enhancers for Noise Cancellation. Aust.
J. Basic Appl. Sci. 2012, 6, 337–352.
91
Appl. Sci. 2019, 9, 294
16. Murano, K.; Unagami, S.; Amano, F. Echo Cancellation and Applications. IEEE Commun. Mag. 1990, 28,
49–55. [CrossRef]
17. Messerrschmi, D. Echo Cancellation in Speech and Data Transmission. IEEE J. Sel. Areas Commun. 1984, 2,
18. Bürger, W. Space-Time Adaptive Processing: Fundamentals. Adv. Radar Signal Data Process. 2006, 6, 1–14.
19. Schwark, C.; Cristallini, D. Advanced multipath clutter cancellation in OFDM-based passive radar systems.
In Proceedings of the 2016 IEEE Radar Conference (RadarConf 2016), Philadelphia, PA, USA, 2–6 May 2016.
20. Zhu, Y.; Weight, J.P. Ultrasonic nondestructive evaluation of highly scattering materials using adaptive
filtering and detection. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 1994, 41, 26–33.
21. Monroe, D.; Ahn, I.S.; Lu, Y. Adaptive filtering and target detection for ultrasonic backscattered signal. In
Proceedings of the 2010 IEEE International Conference on Electro/Information Technology, Normal, IL, USA,
20–22 May 2010.
22. Haykin, S. Adaptive Filter Theory, 5th ed.; Pearson: Harlow, UK, 2014.
23. Cioffi, J. Limited-Precision Effects in Adaptive Filtering [invited paper, special issue on adaptive filtering.
IEEE Trans. Circuits Syst. 1987, 34, 821–833. [CrossRef]
24. Sayed, A.H.; Al-Naffouri, T.Y. Mean-square analysis of normalized leaky adaptive filters. In Proceedings of
the 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing (Cat. No.01CH37221),
Salt Lake City, UT, USA, 7–11 May 2001; Volume 6, pp. 3873–3876.
25. Bismor, D.; Pawelczyk, M. Stability Conditions for the Leaky LMS Algorithm Based on Control Theory
Analysis. Arch. Acoust. 2016, 41, 731–739. [CrossRef]
26. Lowe, M.J.S.; Cawley, P. Long Range Guided Wave Inspection Usage—Current Commercial Capabilities and Research
Directions; Imperial College London: London, UK, 2006; Available online: http://www3.imperial.ac.uk/pls/
portallive/docs/1/55745699.PDF (accessed on 20 November 2018).
27. Meitzler, A.H. Mode Coupling Occurring in the Propagation of Elastic Pulses in Wires. J. Acoust. Soc. Am.
1961, 33, 435–445. [CrossRef]
28. Catton, P. Long Range Ultrasonic Guided Waves for Pipelines Inspection; Brunel University: London, UK, 2009.
29. Nurmalia. Mode Conversion of torsional Guided Waves for Pipe Inspection: An electromagnetic Acoustic Transducer
Technique; Osaka University: Osaka Prefecture, Japan, 2013.
30. Nakhli Mahal, N.; Mudge, P.; Nandi, A.K. Comparison of coded excitations in the presence of variable
transducer transfer functions in ultrasonic guided wave testing of pipelines. In Proceedings of the 9th
European Workshop on Structural Health Monitoring, Manchester, UK, 10–13 July 2018.
31. Nakhli Mahal, N.; Muidge, P.; Nandi, A.K. Noise removal using adaptive filtering for ultrasonic guided
wave testing of pipelines. In Proceedings of the 57th Annual British Conference on Non-Destructive Testing,
Nottingham, UK, 10–12 September 2018; pp. 1–9.
32. Wilcox, P.D. A rapid signal processing technique to remove the effect of dispersion from guided wave signals.
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2003, 50, 419–427. [CrossRef]
33. Sanderson, R. A closed form solution method for rapid calculation of guided wave dispersion curves for
pipes. Wave Motion 2015, 53, 40–50. [CrossRef]
34. Wilcox, P.; Lowe, M.; Cawley, P. The effect of dispersion on long-range inspection using ultrasonic guided
35. Guan, R.; Lu, Y.; Duan, W.; Wang, X. Guided waves for damage identification in pipeline structures: A
review. Struct. Control Health Monit. 2017, 24, e2007. [CrossRef]
36. Lowe, M.J.S.; Alleyne, D.N.; Cawley, P. The Mode Conversion of a Guided Wave by a Part-Circumferential
Notch in a Pipe. J. Appl. Mech. 1998, 65, 649–656. [CrossRef]
37. Miao, H.; Huan, Q.; Wang, Q.; Li, F. Excitation and reception of single torsional wave T(0,1) mode in pipes
using face-shear d24 piezoelectric ring array. Smart Mater. Struct. 2017, 26, 025021. [CrossRef]
38. Rose, J.L. Ultrasonic Guided Waves in Solid Media; Cambridge University Press: New York, NY, USA, 2014.
39. Alleyne, D.N.; Pavlakovic, B.; Lowe, M.J.S.; Cawley, P. Rapid, Long Range Inspection of Chemical Plant
Pipework Using Guided Waves. Key Eng. Mater. 2004, 270–273, 434–441. [CrossRef]
40. Douglas, S.C. Performance Comparison of Two Implementations of the Leaky LMS Adaptive Filter. IEEE
Trans. Signal Process. 1997, 45, 2125–2129. [CrossRef]
41. Cartes, D.A.; Ray, L.R.; Collier, R.D. Experimental evaluation of leaky least-mean-square algorithms for active
noise reduction in communication headsets. J. Acoust. Soc. Am. 2002, 111, 1758–1771. [CrossRef] [PubMed]
92
Appl. Sci. 2019, 9, 294
42. Kuo, S.M.; Lee, B.H.; Tian, W. Real-Time Digital Signal Processing: Implementations and Applications; John Wiley
& Sons: Chichester, UK, 2006.
43. Ssanalysis. Abaqus/Explicit. Available online: http://www.ssanalysis.co.uk/ (accessed on 20 Nov 2018).
44. Teletestndt. Long Range Guided Wave Testing with Teletest Focus+. 2018. Available online: https:
//www.teletestndt.com/ (accessed on 08 January 2019).
45. Alleyne, D.N.; Lowe, M.J.S.; Cawley, P. The Reflection of Guided Waves from Circumferential Notches in
Pipes. J. Appl. Mech. 1998, 65, 635–641. [CrossRef]
46. Lowe, P.S.; Sanderson, R.M.; Boulgouris, N.V.; Haig, A.G. Inspection of Cylindrical Structures Using the First
Longitudinal Guided Wave Mode in Isolation for Higher Flaw Sensitivity. IEEE Sens. J. 2016, 16, 706–714.
[CrossRef]
47. Niu, X.; Duan, W.; Chen, H.; Marques, H.R. Excitation and propagation of torsional T (0, 1) mode for guided
wave testing of pipeline integrity Excitation and propagation of torsional T (0, 1) mode for guided wave
testing of pipeline integrity. Measurement 2018, 131, 341–348. [CrossRef]
48. Duan, W.; Niu, X.; Gan, T.; Kanfound, J.; Chen, H.-P. A Numerical Study on the Excitation of Guided Waves.
Metals 2017, 7, 552. [CrossRef]
50. Alleyne, D.N.; Cawley, P. The excitation of Lamb waves in pipes using dry-coupled piezoelectric transducers.
J. Nondestruct. Eval. 1996, 15, 11–20. [CrossRef]
51. Engineer, B.A. The Mechanical and Resonant Behaviour of a Dry Coupled Thickness-Shear PZT Transducer
Used for Guided Wave Testing in Pipe Line. Ph.D Thesis, Brunel University, London, UK, 2013.
52. Mathworks. Matlab. Available online: https://uk.mathworks.com/products/matlab.html (accessed on
18 April 2018).
93
applied
sciences
Article
Defect Detection using Power Spectrum of Torsional
Waves in Guided-Wave Inspection of Pipelines
Received: 24 January 2019; Accepted: 1 April 2019; Published: 6 April 2019
Abstract: Ultrasonic Guided-wave (UGW) testing of pipelines allows long-range assessment of

pipe integrity from a single point of inspection. This technology uses a number of arrays of
transducers separated by a distance from each other to generate a single axisymmetric (torsional)
wave mode. The location of anomalies in the pipe is determined by inspectors using the received
signal. Guided-waves are multimodal and dispersive. In practical tests, nonaxisymmetric waves are
also received due to the nonideal testing conditions, such as presence of variable transfer function of
transducers. These waves are considered as the main source of noise in the guided-wave inspection
of pipelines. In this paper, we propose a method to exploit the differences in the power spectrum of
the torsional wave and flexural waves, in order to detect the torsional wave, leading to the defect
location. The method is based on a sliding moving window, where in each iteration the signals
are normalised and their power spectra are calculated. Each power spectrum is compared with
the previously known spectrum of excitation sequence. Five binary conditions are defined; all of
these need to be met in order for a window to be marked as defect signal. This method is validated
using a synthesised test case generated by a Finite Element Model (FEM) as well as real test data
gathered from laboratory trials. In laboratory trials, three different pipes with defects sizes of 4%,
3% and 2% cross-sectional area (CSA) material loss were evaluated. In order to find the optimum
frequency, the varying excitation frequency of 30 to 50 kHz (in steps of 2 kHz) were used. The results
demonstrate the capability of this algorithm in detecting torsional waves with low signal-to-noise
ratio (SNR) without requiring any change in the excitation sequence. This can help inspectors by
validating the frequency response of the received sequence and give more confidence in the detection
of defects in guided-wave testing of pipelines.
Keywords: signal processing; defect detection; torsional wave; power spectrum; sliding window;
pipeline inspection; ultrasonic guided-waves (UGWs)
1. Introduction
Pipelines are the main means of transferring oil and gas. They are usually installed in hostile
locations and must be inspected to avoid failures that would be harmful to the environment. Ultrasonic
guided-wave (UGW) enables long-range inspection of pipelines from one single test point. For example,
the UGW technique in pulse-echo mode can enable inspections of up to 50 m (in normal testing
conditions); hence, it is cost efficient in the inspection of large structures. The received signals
are inspected in time-domain where inspectors spot anomalies in the data based on their local
signal-to-noise ratio (SNR). In any region where the envelope of the signal has higher energy than
the regional noise level, an anomaly is reported which can indicate a defect or a feature of the
pipe [1–3]. UGWs are multimodal: for each excitation frequency, multiple wave modes are transmitted

Appl. Sci. 2019, 9, 1449
and received. In the inspection, the goal is to achieve a pure axisymmetric wave, but due to the
imperfect testing conditions [4,5], nonaxisymmetric waves will also be received. This is the main
source of the coherent noise in the inspection. To allow ease of inspection, many researchers have
applied digital signal processing methods that use the differences in these wave modes to either
detect the axisymmetric wave or filter the nonaxisymmetric waves to increase the SNR of the received
anomalies. One of the widely used techniques in the signal processing of guided-waves is the work
of Wilcox et al. [6], which uses previously calculated dispersion curves in order to compensate for
the effect of dispersion for the wave mode of interest in a certain propagation distance. In 2013,
Zeng et al. [7] used the basis of the dispersion compensation technique and introduced a novel method
to design waveforms in order to precompensate for a certain propagation distance.
Pulse compression is another approach that has been previously investigated in the literature.
Instead of exciting a narrowband sine wave sequence, pulse compression uses a known coded
sequence and processes the received signal by applying match-filtering in order to detect the known
sequence. One of the initial attempts of testing pulse compression in the guided-wave was done by
Rodriguez et al. [8,9], where chirp sequences were excited using air-coupled piezoelectric transducer
in order to generate Lamb waves in aluminum plates. Higher SNR and peak values for the signals of
interests were achieved in the experiments; however, the effect of dispersion was not considered, and
therefore a decrease in signal amplitude was expected. In 2010, they published another paper using the
same system where the phase modulation based on Golay codes were used where the enhancement of
21 dB in SNR using a 16-bit Golay code was achieved compared to the conventional pulse transmission.
Mehmet et al. [10,11], introduced an iterative dispersion compensation for removing the dispersion
effect of guided-wave propagation. Most recently, Malo et al. [12] developed a two-dimensional
compressed pulse analysis in order to enhance the achieved SNR of each wave mode. In the pulse
compression approaches, the results are always a trade-off between the spatial resolution and SNR,
as having a result with higher propagation energy and δ-like correlation means that the signal duration
must be increased. Furthermore, the transfer function of the excitation system can affect the accuracy
of the coded waveforms. These two factors indicate that the initial waveform design and the accuracy
of inspection using this method depend on the testing conditions.
On the other hand, other approaches have also been reported in the literature where
narrowband sine waves were used as excitation sequence. Kamran et al. investigated split-spectrum
processing [13–15], where the time-domain signal is decomposed into multiple signals in different
frequency bands, and then recombined in time-domain in order to remove the coherent noise.
In their most recent work [15], it was reported that using the optimum filter parameters and polarity
thresholding method for recombination, the SNR of defects with sizes as small as 2% cross-sectional
area (CSA) could be improved significantly. Nonetheless, the technique depends on various parameters
which can depend on the pipes’ characteristics. Another recently developed method is the spectral
subtraction, investigated by Duan et al. [16]. In this method, the noise signature is calculated using
a small section of the retrieved signal where no real pipe feature exists. Afterwards, this signal is
subtracted from the total signal using a sliding window where a significant reduction of coherent noise
level could be achieved. Nonetheless, achieving noise signature in the practical inspection of pipelines
is difficult as the location of defects and even pipe features might be unknown.
Guided-waves are dispersive and multimodal [17]. Depending on the excitation frequency,
multiple wave modes can be generated at the point of excitation. Guided-wave modes are categorised
based on their displacement patterns (mode shapes) within the structure. Three main families of waves
exist in pipes, which are longitudinal, torsional and flexural. Longitudinal and torsional waves are
axisymmetric waves, while flexural waves are nonaxisymmetric. The popular nomenclature used
for them are in the format of X (n,m), where X can be replaced by the letters L for longitudinal, T for
torsional and F for flexural waves; n shows the harmonic variations of displacement and stress around
the circumference and m represents the order of existence of the wave mode [18]. In general inspection,
the practice is to generate a single pure axisymmetric wave which tends to ease the interpretation of
96
Appl. Sci. 2019, 9, 1449
results. However, variations in the transducers’ transfer function and their placement will cause the
flexural waves to be received; furthermore, wave mode conversion also causes flexural waves to be
received [5,17–19].
Dispersion causes the energy of a signal to spread out in space and time as it propagates [6].
Figure 1a shows an example of a dispersion curve calculated from 8-inch schedule 40 steel pipe using
RAPID software [21]. Wilcox et al. introduced a method to use dispersion curves in order to both
simulate [22] and remove [6] the effect of dispersion. Using the developed formula, an example of
a dispersive wave is shown in Figure 1b, which is based on the dispersion curves of F (4,2). Unlike
T (0,1), which is nondispersive across its whole frequency range [23], flexural waves are dispersive.
Therefore, for ease of inspection, the torsional wave which is both axisymmetric and nondispersive
is used.
7 ) ) )

) ) )

3KDVH9HORFLW\PV

)UHTXHQF\N+]
(a) (b)
Figure 1. (a) Example dispersion curve of T(0,1) wave mode in an 8” schedule 40 steel pipe. (b) The
effect of dispersion on a simulated flexural wave for two propagation distances [20].
The common concept in all the aforementioned signal processing methods is the usage of
dispersion and multimodality of guided-waves in order to increase the SNR. Afterwards, the signals
are inspected in time-domain to report on the location of anomalies. It has already been demonstrated
in the literature that the spectral domain can be processed to remove the noise. Looking at the problem
with another perspective, instead of noise removal, the spectral domain can also be used to detect the
signal of interests (defect) automatically. The excitation sequence is not dispersive and axisymmetric;
hence, no significant changes in the frequency response of the received signal should be inspected.
Therefore, a sliding window is used where the power spectrum of each iteration is compared with the
one achieved from the excitation sequence.
The primary motivation is to enable automated inspection of pipelines without the need for any
inspectors. However, at this stage, the developed algorithm can be used as a tool, along with the
conventional detection methods, to provide more certainty in interoperation of the results and reduce
the number of outliers called due to the coherent noise. It is expected that experience of this algorithm
will help develop automated inspection of pipelines. It should be noted that the experiments were
performed with shorter pipe length. Nevertheless, the detectability of the defects is mostly dependent
on the characteristics of the received signal. The algorithm can easily detect a perfect axisymmetric
feature such as a weld in long-range. Nonetheless, for defects which are generally smaller in size
and asymmetric, the hypothesis is that the feature can be detected if the characteristic of the torsional
wave is still detectable in the received signal. It should also be noted, that as the inspection pipe
length increases the energy of the flexural waves reduce due to their dispersion in time and the effect
of attenuation on them. Therefore, false alarms using this method would be expected to be less in
long-range inspections.
97
Appl. Sci. 2019, 9, 1449
The paper is organised as follows. Section 2 explains the proposed method for detection of defects
using torsional waves. Section 3 demonstrates the results of the method on both synthesised and real
experimental data and, finally, Section 4 concludes the paper.
2. Methodology
In ultrasonic testing, the excitation sequence is known, and the inspectors generally look for
signals with the same characteristics in the time domain to find their signals of interest, such as a defect
in materials. In UGW, one of the common excitation sequences is a 10-cycle Hann windowed signal
with different frequency, depending on the testing conditions. However, these signals are typically
polluted by coherent noise which has similar temporal characteristics to those of defect signals.
It was mentioned beforehand that in the guided-wave, torsional waves are nondispersive and
axisymmetric. The flexurals, which are the main source of coherent noise in the tests, are dispersive and
nonaxisymmetric. In the power spectrum of an inspected pure torsional wave, the same characteristics
of the spectrum of the excitation sequence must be observed; this is not the case for flexural waves due
to their temporal and spatial variances. It should also be mentioned that asymmetric features (which is
the case for most defects) generally reflect both torsional and flexural waves; where their corresponding
reflection coefficients depends on the geometry of the feature. However, in this research, we focus
our detection based on the detection of the torsional waves. The torsional wave is always reflected,
and it is received symmetrically around the pipe circumference regardless of the features’ geometry
or the wave propagation distance. Furthermore, in some cases, the flexural waves are considered as
the coherent noise received due to mode conversion from the known features, leading to detection of
false alarms in the tests. Hence, the focus is in detection of torsional wave in order to find the location
of features.
In this paper, we propose a condition-based comparison of the power spectrum achieved from a
moving window of the received signal and the spectrum of the normal excitation sequence. Keeping
in mind that, since both flexural and torsional waves have the same main bandwidth (BW), correlating
them would not be sufficient for distinguishing between them. Nonetheless, using this proposed
method the torsional wave is identifiable. Another advantage of this method is that it is relying on the
excitation sequence, which is known and set manually by the inspectors.
This algorithm consists of three main aspects: (1) initialisation, which initialises the excitation
sequence and extract the required features for comparison; (2) main loop, that uses the advancing
window and carries out the pre- and postprocessing of the conditions; and (3) conditions, which
constitute the main processing, where the spectrum of each iteration is compared with the one
achieved from excitation sequence.
Since the signals are processed digitally, and due to the limited resolution in this domain, in the
following section all formulae and definitions are presented based on the sample number for the
time-domain and bin number for the power spectrum. In doing so, better performance, in terms of
speed is achieved as no unnecessary interpolation, is needed. The sampling frequency (Fs) used in the
tests is 1 MHz which is a fairly common sampling frequency of guided-wave inspection devices.
2.1. Initialisation
Initially, the excitation sequence is created where the inspector inputs the centre frequency and the
number of cycles. Then, the “windowSize” is calculated based on the number of samples required for
the excitation sequence. This sequence is already normalised by its maximum value. The normalisation
bounds the signal amplitudes between 1 and −1 by dividing each sample by the maximum amplitude
within the signal. Since 10 cycles are typically used for inspection, the duration of the excitation
sequence is small; therefore, this signal is zero padded by a factor of 4 to achieve better resolution in
the spectral density. Zero padding is the operation of adding extra zeroes to the end of the sequence in
order to increase the resolution of the power spectrum [24]. The power spectrum is then evaluated by
98
Appl. Sci. 2019, 9, 1449
applying the fast Fourier transform [24], where the magnitudes of the spectrum are saved in “signalRef ”
and the corresponding frequency for each bin is saved in “frqList”.
The first extracted feature from the spectrum is the maximum magnitude and its corresponding
bin number which are stored in “maxRef ” and “idFC”, respectively. This should be the bin representing
the centre frequency of excitation as set before by the inspector. Afterwards, the lowest and highest
frequencies of the 10 dB BW of this spectrum are calculated. The bin numbers of the lowest and highest
frequencies within this BW are saved in “teFLID” and “teFHID”. If the previously calculated “idFC”
is not in the centre of this spectrum, these numbers are expanded away from the centre frequency
so both the lower half and the upper half of the spectrum have the same amount of bin numbers.
Then, the greater magnitude achieved from these (border) bins is stored in “teMax”. The flowchart
of this function is shown in Figure 2 and a summary of the extracted variables are shown in Table 1.
All values except “windowSize” are needed for the conditions function.
Figure 2. Flowchart of the initialisation.
Table 1. Description of variables extracted from the Initialisation.
Variable Name Description Conditions

windowSize Length of the moving window -
signalRef The power spectrum of the normalised excitation sequence C2, C3, C4
frqList List of the corresponding frequency for each bin number C0
maxRef Maximum magnitude achieved from signalRef C4
idFC Bin number of maxRef C0
teFLID Bin number of the lowest frequency within 10 dB bandwidth C0, C1, C3
teFHID Bin number of the highest frequency within 10 dB bandwidth C0, C1, C3
teMax The greater magnitude between teFLID and teFHID from signalRef C3
99
Appl. Sci. 2019, 9, 1449
2.2. Conditions
In this function, all previously extracted characteristics are used in order to compare the spectrum
of the moving window with the excitation sequence. The focus of this paper is on the similarity of
the spectrum rather than the energy of the signals; thus, all signals are normalised before calculating
the spectrum. Five conditions (features) exist in total. Since defect signals are affected by the flexural
noises, it is expected that the received signals from smaller defects would not result in a perfectly
matching spectrum. Hence, in each condition, safety variables are defined in order not to over-fit the
conditions to the reference. These variables are set by the inspectors to provide a safety tolerance
for detecting weak signals. It should be borne in mind that these tolerances have typically small
values and can be fixed. Furthermore, they are set only once at the start of the algorithm and does not
change iteratively.
2.2.1. Condition Zero (C0)

The centre frequency of the windowed spectrum must be near the centre frequency of the
excitation sequence, defined by “idFC”. In case of small defect signals, where less energy is reflected,
they are more affected by the coherent noise which will cause a slight shift in the centre frequency.
Therefore, a slight shift in the frequency is acceptable which is defined by variable “maxFCallow”.
If such a shift is detected, the corresponding bin numbers of “idFC”, “teFLID” and “teFHID”, which are
required for the other conditions, will be shifted towards the new centre frequency of the spectrum.
If this shift of centre frequency is more than maximum allowable shift, it means that the current iteration
is significantly different to the excitation sequence and all other conditions will automatically fail.
The maximum allowable shift, “maxFCAllow” is set to 3 kHz, which is a small margin in comparison to
the total BW of the signal.
2.2.2. Condition One (C1)

In the Hann windowed sine waves, the lower half of the BW tends to increase in magnitude
when moving towards the centre frequency then it starts to decrease. In this condition, it is checked
that there is no discontinuity in this trend; in other words that no local maxima exist. In doing so,
the 10 dB BW is of the main focus as the outside edges are typically affected by the flexural noises due
to their lesser magnitudes. The reason for this is that in guided-wave inspection, (Hann) windowed,
narrowband excitation sequences are used. Therefore, the main energy spread is within the main BW
of the signal while a limited amount will be assigned to the borders of BW. The flexurals are dispersive,
depending on their dispersion curves, their spectra tend to shift towards other frequencies, but the
nondispersive torsional wave will not observe this change. Therefore, the energy of flexural waves
will be stronger outside the main BW. Choosing 10 dB BW is a good compromise since more than 90%
of the bins is covered using this value and only a few bins on the sides are neglected. Using the bin
number of lowest frequency and the highest frequency in the 10 dB BW, “teFLID” and “teFHID”, it is
checked whether the magnitude of each respective bin is increasing when moving toward the centre
frequency. Nonetheless, a variable can also be defined as “diffVal” to allow small differences to be
neglected. Currently, no tolerance is needed, and it is kept as zero.
2.2.3. Condition Two (C2)

In the excitation sequence, the strongest magnitudes are detected in the region of the centre
frequency. Since they have the strongest magnitude, they will be less affected by the noise in the region.
In this condition, it is confirmed that the neighbourhood of the centre frequency has relatively the
same magnitude. The number of neighbourhoods in each side of the centre frequency is defined by
“cBins“, and the maximum allowable difference is defined based on a percentage of current iteration
maximum magnitude defined by “maxPerc”. In this paper, the number of the neighbourhood is set as 2
and the maximum allowable difference is set as 20% of the maximum magnitude of each iteration.
100
Appl. Sci. 2019, 9, 1449
2.2.4. Condition Three (C3)

The window size is set as the length of the excitation sequence. Due to this limited duration of the
signal, in the iteration’s power spectrum, there should not be any magnitudes greater than the 10 dB
BW boundary, “teMax”, of the excitation sequence other than the frequencies within these boundaries.
Nonetheless, in the immediate bins next to the boundaries, “teFLID” and “teFHID”, the magnitudes
are more affected by the coherent noise and can be greater than the “teMax”. Therefore, “sBins” is
defined as the allowable number of immediate bins, which will be neglected during this comparison.
This value is set to 1.
2.2.5. Condition Four (C4)

Each iteration window is normalised before its power spectrum is calculated. As the window size
is limited to the excitation sequence, in the power spectrum of the iteration, a loss of magnitude can be
expected due to the coherent noise in the signals. However, the maximum magnitude of the iteration
cannot be greater than the one achieved from the excitation sequence, “maxRef ”. Since normalisation
is taking place, a small safety margin is defined as “sMargin”, to compensate for the small variances
between the magnitudes defined. This value is set to 1, which is much less than the maximum
magnitudes achieved from excitation sequences.
2.2.6. Final Results

Iterations are marked as the signal of interest (one) if all the aforementioned conditions are met
and will be marked as noise (zero) if any of them are failed. Furthermore, the maximum correlation
between the 10 dB BW of the power spectrum of the excitation sequence and current iteration is
calculated. The cross-correlation of two sequences x[n] and y[n] is defined as [24]
N
rxy [l] = ∑1 y[n + l]x[n] (1)
where l is the (time) shift (or) lag, n is the sample number of each sequence and N is the number of
samples in the sequences. The correlation operation measures the degree to which two sequences
are similar. As it can be seen in the formula, the output is given in a vector of time shifts. Therefore,
by considering the maximum amplitude of rxy [l], the maximum similarity between two sequences is
measured. Nonetheless, since in C0, the peaks are shifted, no time shifting is required to achieve the
maximum value and the results will become the summation of multiplication of each corresponding
bin. Both the cross-correlation value and the binary condition results (“Total”) will be returned to the
main loop where the final “detection” results will be generated. The flowchart of this function is shown
in Figure 3.
101
Appl. Sci. 2019, 9, 1449
Figure 3. Flowchart of conditions.
2.3. Main Loop

The initialisation is done before the main loop in order to extract the required features. Afterwards,
the start of the iteration is delayed by the length of the window size. In practice, it means that the
system can be linear and time invariant, and can be implemented in real-time as it only requires past
samples of the signals. In each iteration, a temporary signal, “WindowedSignal”, is created which holds
102
Appl. Sci. 2019, 9, 1449
the past samples of the signals with the same length as the excitation sequence. Since the power
spectra are needed to be compared, the signals are first normalised and zero padded by a factor of 4.
The power spectrum of this windowed signal is calculated and passed to the conditions function. After
processing, the “Results” object is returned where it contains the detection variable “Total” and the
correlation value of the spectrums as “Value”. If the signal of interest is detected, e.g., “Total” is equal to
one, then the current iteration represents a similar power spectrum to the one of excitation sequence.
One approach would be to mark the current iteration number as one which would mean that
the current iteration is representing the signal of interest; however, this is not the best representation
as the current iteration might not be the best representative of the windowed signal characteristics.
In this paper, the index of the maximum amplitude in each windowed signal is detected. This index
appears to be a better representation since the region with more energy has a stronger influence on
the power spectrum. Since the excitation sequence is Hann windowed, the centre of the signal will
hold the highest amount of signals’ energy. This as in this location, the concentration of signal energy
is higher, stronger flexural features are required to disturb reduce/change the characteristics of this
region. Therefore, the centre instead of adding the correlation value to the sides of the window (current
iteration number), the value is cumulatively added to the location of the highest amplitude within the
window, which represents the centre of the Hann windowed sequence in a torsional wave. This value
is stored in “detection” vector. Therefore, the correlation value is added to the location where the signal
represents the strongest amplitude in the “detection” vector. The reason why they are added rather
than replaced is that since it is an iterative process, different windows can have the same index as the
maximum; by doing a cumulative sum for each index, the certainty in defect detection is increased and
the amplitudes assigned for the outliers and noises are decreased. The main loop finishes when the
end of the signal is reached, at which point the “detection” vector is normalised and is plotted against
the original signal to show the defect locations and their normalised correlation values. The flow chart
of this function is shown in Figure 4.
Figure 4. Flowchart of the main loop.
103
Appl. Sci. 2019, 9, 1449
3. Results
This method was initially developed on the FEM test case. As explained previously, the FEM
model considers both spatial and temporal characteristics of guided-waves in order to simulate a
coloured noise with similar characteristics of the noise in the real inspection. For a better understanding
of the methodology, each condition is firstly explained using this FEM signal. To validate this method,
the results on experimental pipes are shown. SNRs reported in this section as follows
⎛ ⎞
1 Signal[n]
∑N
1
N
SNR = 20 ∗ log 10⎝ ⎠ (2)
1 Noise[m]
∑M
1
M
where “Signal” is considered as the samples where the defect (torsional wave) is expected to be received
and every other sample is considered as Noise. In all tests, the excitation sequences are of 10-cycle
Hann windowed sine waves, which are commonly used in the field inspections.
3.1. FEM Test Case

The generated signal in the FEM test case has the excitation frequency of 30 kHz, as shown
in Figure 5. This setup of the model was previously explained in [25]. Modelling of guided-wave
signals using FEM have been widely used and validated in the literature [2,26–28]. However, the
major difference in this model is that the excitation sequences have variable transfer functions and
fewer reception points are used; this tends to add flexural noise in data. In this figure, for better
illustration of the signal, the received signal from (forward) the pipe end and the starting dead zone
are removed. The defect signal is expected to be received at 3–3.5 m. Since highly variable transfer
functions have been used, the backward cancellation algorithm is not able to cancel the backward
going signal perfectly. This reversed wave reflected from the pipe end is received at 1–1.5 m. Due to
this backward leakage, another signal of interest, which is a torsional wave, is expected to be received
at approximately 4–4.5 m. This signal is generated when the echo of the backward leakage and the
defect signal (in the forward testing direction) is received at the test tool. All other regions are the
noise caused by the flexural waves.
Figure 5. Synthesised data generated from finite element model (FEM) test case.
104
Appl. Sci. 2019, 9, 1449
3.1.1. Condition Zero

The starting condition is that the maximum magnitude of each iteration’s power spectrum must
belong to the approximate region of the centre frequency. This approximation is set by the inspector
(“maxFCAllow”). It is a necessary factor for the detection of smaller defects since the spectrum will
be more affected by the flexural waves, which leads to a shift in centre frequency. An example of an
outlier based on this condition is shown in Figure 6, where the blue box shows an example allowable
region for frequency shift, and the grey circle shows the bin with the maximum magnitude in this
iteration. Since it is outside the allowable region, the outcome of this condition will be zero, which will
automatically force all other conditions to zero. The used value for the maximum allowable shift is
3 kHz which is a small margin in comparison to the overall BW of the excitation sequence. In the case
of FEM signal, since the flexural wave does not affect the defect significantly, similar results can be
achieved even when this value is set to zero.
Figure 6. Example of an outlier case detected using Condition Zero (C0), where (a) shows the time
domain of the iteration window and (b) is its respective power spectrum. The red lines (dotted, +)
show the references achieved from excitation sequence and the black lines (x, solid) show the results
from each iteration.
3.1.2. Condition One

In simple terms, the magnitudes of each bin in the 10 dB BW of the iterations’ spectra must
increase when moving toward the centre frequency. Due to dispersive nature and multimodality of
guided-waves, this condition is not true for many iterations as a combination of waves will be received
with different centre frequencies. Nonetheless, this condition must be true in the case of the torsional
wave as it will have the same characteristic of the excitation waveform. Figure 7 shows an example
of an outlier detected based on this condition. As can be seen, in the region between 26.5 to 27.5
kHz, the magnitude of three frequencies are actually decreasing, and a local minima is created. This
suggests that the iteration is actually two separate wave modes with two different centre frequencies.
105
Appl. Sci. 2019, 9, 1449
Figure 7. Example of an outlier case detected using Condition One (C1), where (a) shows the time
domain of the iteration window and (b) is its respective power spectrum. The red lines show (dotted,
+) the references achieved from excitation sequence and the black lines (solid, x) show the results from
each iteration.
In this condition, in order to avoid marginal errors due to the limited number of samples in the
window, a safety variable (“diffValue”), so the condition holds true if a slight decrease in the magnitude
of each consecutive bin, is detected. In these tests, this value is kept at zero which means strictly positive
increase of the magnitudes must be observed when moving toward centre frequency. The reason is, in
case of guided-waves testing, if the flexural waves are strong enough that the frequency characteristics
of the excitation waveform are changed, the signal must not be marked as a defect; since, in effect, the
iteration is representing a strong existnace of a flexural wave.
3.1.3. Condition Two

The strongest magnitudes are in the neighbourhood area of the centre frequency; hence, they are
less affected by the flexural waves. This condition verifies that the magnitudes from centre frequencies
are not below a certain limit. Figure 8 shows an example of an outlier detected from this condition.
Two limits must be set for this condition: (1) the number of neighbouring in “cBins” and (2) the
percentage of allowable drop with regards to the maximum magnitude detected in the iteration
(“maxPerc”). In these tests, “cBins” is set as two (of each side of centre frequency), which means a
total of four bins is checked and thus limits this validation to the main central bins of the spectrum.
In the reference signal, it can be seen that at the border of this region, an approximate 10% drop is
observed. Since in practical testing, the magnitudes will be more affected by the flexural waves and
testing condition, the allowable drop is set to 20% of the maximum magnitude of each iteration.
106
Appl. Sci. 2019, 9, 1449
Figure 8. Example of an outlier case detected using Condition Two (C2), where (a) shows the time
show the references achieved from excitation sequence and the black lines (solid, x) show the results
3.1.4. Condition Three

In the third condition, the magnitudes of frequencies outside the 10 dB BW of the excitation
sequence are checked. Since the window size is limited to the length of the excitation sequence,
no other strong peaks in other frequencies should be detected. This can be verified by confirming that
no other magnitudes are above the lowest magnitude within 10 dB BW. Figure 9 shows an example
for an outlier detected using this condition. The blue boxes illustrate the expected regions while the
grey boxes illustrate the bins with higher magnitudes than the thresholds. It should be noted that this
threshold is defined by the reference signal rather than the excitation sequence. Since the boundaries
are affected more by flexural waves, a safe margin can be provided by expanding the bins away from
the centre frequency (“sBins”). In these tests, this value is defined as one, which means only one bin
is neglected from each side of the spectrum and all other bins must have magnitudes less than the
minimum magnitude detected in the 10 dB BW.
107
Appl. Sci. 2019, 9, 1449
Figure 9. Example of an outlier case detected using Condition Three (C3), where (a) shows the time
3.1.5. Condition Four

In this condition, it is verified that the magnitude achieved in the power spectrum of iteration
is not more than the reference. The reference signal is in the perfect form of the excitation when it
is normalised and not affected by any noise, while any the windowed iterations are subject to noise;
hence, the maximum magnitude of the iteration cannot be more than the reference. Nonetheless, due to
the finite resolution of the windows, a safety value is defined as “sMargin” to allow small differences
between the two to be neglected.
Keeping in mind that in the case of FEM signal, since source points were not placed linearly and
variable transfer functions were used, the characteristics of the generated signals can be different.
This is clearly illustrated in Figure 10, where the received has more than 10 cycles; hence, the maximum
magnitude in the power spectrum of the iteration would be increased. The shown windowed signal
is approximately located around the 4-metre region (second torsional wave), where the results of
impact from backward leakage and the front defect are received from the forward direction. As can
be seen, the condition will not hold in this case; however, when setting the threshold as 10% of the
maximum magnitude (15) the result of the first detection is increased and this wave will also be
detected. Changing this value to 15% increases the detection amplitude of the second signal and
the pipe end even more, but the first defect signal remains constant. At 20%, the maximum values
are achieved (while the overall detection result of the defect becomes less due to the increment of
the associated value to the pipe end), and anything above does not change the generated results.
Nonetheless, for the purpose of experimental tests, this value is only set as one which is less even than
1% of the maximum magnitude of the reference. This ensures that only calculation errors due to the
limitation of the window size are neglected.
108
Appl. Sci. 2019, 9, 1449
Figure 10. Example of an outlier case detected using Condition Four (C4), where (a) shows the time
3.1.6. Results Total

The correlation results of the reference and the iteration’s power spectrum is added to the location
of maximum voltage in each window. Furthermore, as the correlation result is typically a large number,
the final result is normalised by the maximum value, which is typically the pipe end. This is illustrated
in Figure 11 where it can be seen that the value of 0.1 is associated exactly to the centre of the defect
signal. As illustrated, knowing only the excitation waveform, the approximate location of the defect is
clearly detectable using this method.
Figure 11. The final result (red line) overlaid on the time-domain signal (black line) from the FEM Case.
The defect size is 3% cross-sectional area (CSA) and the excitation frequency is 30 kHz.
109
Appl. Sci. 2019, 9, 1449
The binary detection results of each condition for each iteration, as well as the total results,
is demonstrated in Figure 12. In this figure, it is clearly demonstrated outliers exist in the results of
each individual condition, but by combining all of them, only the iterations associated with the defect
and pipe end are being detected. Furthermore, high correlation results are achieved from all iterations,
irrespective of it being noise or defect; which is the main reason why doing a cross-correlation is not
enough for detecting the defect location in the case of guided-wave testing. However, even with the
existence of the coloured-noise, by checking the mentioned condition, the detection becomes possible.
Figure 12. The generated results from the FEM test case.
110
Appl. Sci. 2019, 9, 1449
In the same format, Figures 13 and 14 show the results achieved from the 3% CSA defect from the
experimental results. The same aforementioned conclusions can be drawn with regard to Figure 13.
However, it can be seen that in the experimental results since the transfer functions are more linear,
the outcome of condition 4 (C4) is true in most iterations. The results are shown in Figure 14, where
the defect location is clearly marked using this algorithm. In overall, conditions one and three are the
most effective ones which filter out most of the outliers from the tests.
Figure 13. The generated results from the experimental pipe with a defect of 3% CSA and testing
frequency of 38 kHz.
These five conditions have not been optimised. Nevertheless, they were all designed based on the
physical phenomena of guided-waves propagation within pipes. As can be seen in both the results
of FEA and experimental tests, only in C3 and C4 could the defects be detected. However, having
more relevant conditions, especially for practical inspection of pipes, may increase the certainty of the
generated results. Nonetheless, the conditions are not necessarily required to be in the same Fourier
Domain as stated in this algorithm, and the technique can be combined with other approaches, such as
Wiener filtering, in order to improve its performance.
111
Appl. Sci. 2019, 9, 1449
Figure 14. The final result (red line) overlaid on the time-domain signal (black line) from experimental
test case with the excitation frequency of 38 kHz and defect size of 3% CSA.
3.2. Experimental Results

A Teletest MK4 device [29], manufactured by Teletest branch of Eddify technology family in
Cambridge (UK), was used in order to inspect an eight-inch schedule 40 steel pipe with a total length
of six metres. The test tool included three rings of 24 linearly spaced thickness-shear (d15) transducers
with 30-mm spacing, which are located 1.5 m away from the back end of the pipe. The device is
capable of operating in pulse-echo mode with the operating frequency of 20 to 100 kHz, and maximum
sampling frequency of 1 MHz. On the reception side, the individual sections received from each
transducer is postprocessed in order to reduce the effect of echoes received from backward testing
direction and flexural waves [5,17,30]. The schematic of the signal reception routes with this setup is
shown in Figure 15.
Figure 15. Torsional mode reception route for laboratory trials.
The algorithm was assessed using data gathered from real pipes with three different defect size of
2%, 3% and 4% CSA. The defects were introduced using a saw-cut by thinning the wall thickness of the
pipe’s circumference. The schematics of the defects and their approximate sizes are shown respectively
in Figure 16 and Table 2.
112
Appl. Sci. 2019, 9, 1449
Figure 16. Schematic of the saw-cut defects introduced in the wall of the pipe (not to scale).
Table 2. Defect specifications.
CSA (%) Flaw Depth (mm) Cord Length (mm) Arc Length (mm)
2 3.1 51.75 52.24
3 4.1 59.37 60.12
4 5 65.43 66.48
Depending on the system specification and test specimen, the maximum excitation power
achieved by using frequencies in the range of 34 to 42 kHz. As can be seen in Figure 1a, most
of the created flexurals from conversion of torsional wave is highly dispersive in the lower frequency
region. Since dispersion affects the power spectrum of the received signal, it is expected that this
algorithm works better with lower frequencies. However, the final choice of excitation frequency is
a trade-off between the capability of the system to excite the required waveform and having highly
dispersive flexural waves. Furthermore, it should be noted that the reflected wave amplitude at
different frequencies is generally dependent on the flaw geometry, and this could effect the detection
capability at different frequencies. In order to illustrate this dependency, 10 different frequencies have
been used varying from 30 to 50 kHz with steps of 2 kHz. The achieved SNR from each signal of
experimental tests are demonstrated in Figure 17 and Table 3. As can be seen, 40 kHz achieved the
best SNR in the tests; with a neglectable difference in the case of 2% CSA defect signal in comparison
to other frequency (34 and 42 kHz). Nonetheless, all these defects are either close to the coherent noise
level, which are generated by the unwanted flexural waves or are buried within it. Therefore, detecting
these defects using normal inspection procedures without calling any outliers are challenging.
Table 3. Achieved SNR from defect signal in each test case (in dB).
Frequency (kHz)
CSA (%)
30 32 34 36 38 40 42 44 46 48 50
4 7.13 8.84 11.12 12.15 13.13 13.76 13.23 11.73 9.50 6.84 3.50
3 4.29 5.97 8.15 8.62 9.08 9.51 9.00 7.60 5.43 2.82 −0.51
2 1.46 2.67 3.15 2.29 2.00 2.74 3.09 2.67 0.89 −1.43 −4.44
113
Appl. Sci. 2019, 9, 1449

615'%

)5(48(1&< .+=
Figure 17. Signal-to-noise ratio (SNR) of the defect signal in each experimental test case.
Figure 18 shows the processed results from each of these test cases where (a) shows the amplitude
of detection and (b) shows the amplitude of outlier in the test. Outliers in these tests are considered
as any detection corresponding to regions where the signal from neither defect or pipe features is
expected to be received. In this figure, each line represents a defect with different CSA size and the
value zero represents no detection. Furthermore, it can be seen that the algorithm works best in
frequency range of 34–42 kHz. With these frequencies, defects with sizes above 3% can be detected
with minor or no outliers existing in the results. In lower frequencies, although the amplitude of
outliers is smaller than most test cases, defects with sizes smaller than 4% CSA cannot be detected.
On the other hand, frequencies above 42 kHz generally result in the detection of outliers with higher
amplitudes, where when using 48 and 50 kHz, no defect can be detected. The main reason is the
transfer function of the system, where it is not capable of generating the expected power spectrum for
the excitation waveform. As an example, consider the received signal from 50 kHz test case on 4%
CSA pipe, is shown in Figure 19. The pipe end signal is a pure reflection of the excitation sequence
where limited changes should be observed due to the coherent noise. The retrieved signal from the
pipe end observes a 5 kHz shift in centre frequency. Furthermore, the generated sequence can also
have a greater number of cycles which will result in higher maximum energy in an iteration’s power
spectrum in comparison to the reference. Therefore, because of these significant changes which are
caused by generation of the sequences, it is not recommended to use higher than 42 kHz excitation
frequencies with this algorithm.
Figure 18. Results achieved using the algorithm where (a) shows the detection amplitude of defect
signal and (b) shows the detection amplitude of the outlier. Each line represents a defect with different
CSA size. The red dotted line represents the amplitude threshold for filtering the outliers.
114
Appl. Sci. 2019, 9, 1449
Figure 19. (a) The signal received from pipe end using 50 kHz and (b) its corresponding power
spectrum. The red lines (dotted, +) show the references achieved from excitation sequence and the
black lines (solid, x) show the results from each iteration.
The only frequencies capable of detecting 2% CSA defect are 34, 36 and 40 kHz. With regards to
Figure 20, which demonstrates the defect to outlier detection ratio, the detection amplitude achieved
from defect from 34 and 40 kHz is higher by a factor of 1.5 and 2.38 than the outliers’, while, in the
case of 36 kHz, this factor is only 0.36 which means the amplitude for defect detection is smaller.
Nonetheless, for defects with higher CSA size, the detected defect to outlier amplitude ratio from
tested frequencies in the range of 34 to 42 kHz is almost always higher than a factor of 2. This suggests
that by setting an amplitude threshold, defects above 3% CSA size can be detected without any outliers.
This threshold is shown in Figure 18 by the red dotted line which in this case is set as 0.2. It should also
be borne in mind that doing so, would remove the detection of 2% CSA defect, especially in the case
of 40 kHz since even though the values are smaller than this threshold, but the amplitude of defect
detection (0.19) is approximately twice than those of outlier (0.08). The only two frequencies which
do not detect any outliers in any of the tests are 30 and 38 kHz. Nonetheless, using 38 kHz, 2% CSA
defect and using 30 KHz, both 2% and 3% CSA defects are not detected.
Figure 20. The ratio of detection amplitude of defect to outliers. In cases where the defect is not
detected, the ratio is set as zero.
115
Appl. Sci. 2019, 9, 1449
4. Conclusions
In this paper, a novel method is proposed to detect the location of defects using the known
excitation power spectra of the torsional waves in guided-wave inspection of pipelines. For doing so,
as opposed to traditional inspection using time-domain signal, the spectral domain of the signal is
compared with the characteristics of the known excitation sequence. The method works by applying
a moving window to the received signal from the inspection, where in each iteration the signal
is normalised and its corresponding power spectrum is generated. Since both defect signal and
noise are within the same frequency band, by only measuring the correlation between reference and
iterations power spectrum, the defect signal will not be distinguishable. Therefore, before measuring
the correlation, a total of five conditions were checked to verify that a similar power spectrum to the
reference was achieved:
• Centre frequency shift must be small.
• Considering the 10 dB bandwidth of the iteration’s power spectrum, the magnitude of each
respective frequency bin must be increased when moving toward the centre frequency.
• The achieved magnitude from the neighbouring bins to the centre frequency must be
closely related.
• Frequencies outside the 10 dB bandwidth must have less magnitude than the minimum one
achieved from within the 10 dB bandwidth.
• The maximum magnitude achieved must be less than the one from the reference.
For each condition, a safety margin is set by the user to neglect minor differences due to the
digitisation of the signals and the limitation of the available sample in each window. Nonetheless,
before applying the algorithm on guided-wave data from pipes, these limits can be fixed as they are set
just to provide a tolerance. On the other hand, this method can also be tested in traditional ultrasonic
testing, where there might be a need of tweaking these parameters.
The algorithm was initially developed and validated on synthesised data from FEM. Afterwards,
experimental data form real pipes with defect CSA sizes of 2, 3 and 4%, were used to validate its
capability and also its limitations. It was illustrated that the choice of excitation frequency is a
trade-off between the ability of the system to generate the given excitation sequence and having
highly dispersive flexural waves (which are considered as the main source of the noise in these tests).
As such, it was observed that frequencies in the range of 34 to 40 kHz are optimal, and can detect
defects reliabily with larger CSA sizes than 3% with no or very few outliers. In the case of 2% CSA
defect, only frequencies of 34 and 40 kHz provided robust detection, where the ratio of the defect to
outlier amplitudes is higher by a factor of 1.5. On the other hand, 38 kHz is the only frequency within
this range where no outliers were detected in any of the cases. With regards to the test results, due
to the significant changes in the excitation sequences, it is not recommended to use this algorithm
with frequencies higher than 42 kHz. Since this method only requires a time-domain signal and the
characteristics of the excitation sequence, it can easily be combined with other methods to further
enhance the reliability of defect detection in guided-wave testing of pipelines.
Author Contributions: Conceptualisation, H.N.M.; Data Curation, H.N.M.; Formal Analysis, H.N.M.; Funding
Acquisition, A.K.N.; Investigation, H.N.M.; Methodology, H.N.M.; Project Administration, A.K.N. and H.N.M.;
Resources, H.N.M.; Software, H.N.M.; Supervision, K.Y. and A.K.N.; Validation, H.N.M.; Visualisation, H.N.M.,
K.Y. and A.K.N.; Writing—Original Draft, H.N.M.; Writing—Review & Editing, H.N.M., K.Y. and A.K.N.
Funding: This research study was made possible by the sponsorship and support of Brunel University London
and the National Structural Integrity Research Centre (NSIRC).
References
116
Appl. Sci. 2019, 9, 1449
2. Ostachowicz, W.; Kudela, P.; Krawczuk, M.; Zak, A. Guided Waves in Structures for SHM: The Time-Domain
Spectral Element Method; Wiley: Hoboken, NJ, USA, 2012; ISBN 0470979836.
3. ASTM Standard E2775-16. Standard Practice for Guided Wave Testing of Above Ground Steel Pipeword Using
Piezoelectric Effect Transducer; ASTM: West Conshohocken, PA, USA, 2017.
4. Nakhli Mahal, H.; Mudge, P.; Nandi, A.K. Comparison of coded excitations in the presence of variable
transducer transfer functions in ultrasonic guided wave testing of pipelines. In Proceedings of the 9th
European Workshop on Structural Health Monitoring, Manchester, UK, 10–13 July 2018.
5. Nakhli Mahal, H.; Yang, K.; Nandi, A. Detection of Defects Using Spatial Variances of Guided-Wave Modes
in Testing of Pipes. Appl. Sci. 2018, 8, 2378. [CrossRef]
7. Zeng, L.; Lin, J.; Lei, Y.; Xie, H. Waveform design for high-resolution damage detection using lamb waves.
8. Garcia-Rodriguez, M. Lamb Wave generation with an air-coupled piezoelectric array using square chirp
excitation. In Proceedings of the International Congress on Acoustics, Madrid, Spain, 2–7 September 2007;
pp. 2–7.
9. Garcia-Rodriguez, M.; Yaez, Y.; Garcia-Hernandez, M.J.; Salazar, J.; Turo, A.; Chavez, J.A. Application of
Golay codes to improve the dynamic range in ultrasonic Lamb waves air-coupled systems. NDT E Int. 2010,
43, 677–686. [CrossRef]
10. Yücel, M.K.; Fateri, S.; Legg, M.; Wilkinson, A.; Kappatos, V.; Selcuk, C.; Gan, T.H. Pulse-compression
based iterative time-of-flight extraction of dispersed Ultrasonic Guided Waves. In Proceedings of the 2015
IEEE 13th International Conference on Industrial Informatics (INDIN), Cambridge, UK, 22–24 July 2015;
pp. 809–815.
[CrossRef]
2017, 64, 1092–1101. [CrossRef] [PubMed]
enhancement of ultrasonic guided wave. Ultrasonics 2018, 83, 48–59. [CrossRef] [PubMed]
14. Pedram, S.K.; Haig, A.; Lowe, P.S.; Thornicroft, K.; Gan, L.; Mudge, P. Split-spectrum signal processing for
reduction of the effect of dispersive wave modes in long-range ultrasonic testing. Phys. Procedia 2015, 70,
15. Pedram, S.K.; Mudge, P.; Gan, T.-H. Enhancement of ultrasonic guided wave signals using a split-spectrum
processing method. Appl. Sci. 2018, 8, 1815. [CrossRef]
16. Duan, W.; Kanfoud, J.; Deere, M.; Mudge, P.; Gan, T.-H. Spectral subtraction and enhancement for torsional
waves propagating in coated pipes. NDT E Int. 2018, 100, 55–63. [CrossRef]
17. Lowe, M.J.S.; Cawley, P. Long Range Guided Wave Inspection Usage—Current Commercial Capabilities and Research
Directions; Imperial College London: London, UK, 2006. Available online: http://www3.imperial.ac.uk/pls/
portallive/docs/1/55745699.PDF (accessed on 20 November 2018).
18. Catton, P. Long Range Ultrasonic Guided Waves for Pipelines Inspection; Brunel University: Uxbridge, UK, 2009.
19. Nurmalia. Mode Conversion of Torsional Guided Waves for Pipe Inspection: An Electromagnetic Acoustic Transducer
Technique; Osaka University: Osaka, Japan, 2013.
20. Sanderson, R. A closed form solution method for rapid calculation of guided wave dispersion curves for
pipes. Wave Motion 2015, 53, 40–50. [CrossRef]
22. Guan, R.; Lu, Y.; Duan, W.; Wang, X. Guided waves for damage identification in pipeline structures: A review.
Struct. Control Health Monit. 2017, 1–17. [CrossRef]
23. Nakhli Mahal, H.; Mudge, P.; Nandi, A.K. Noise removal using adaptive filtering for ultrasonic guided
wave testing of pipelines. In Proceedings of the 57th Annual British Conference on Non-Destructive Testing,
Nottingham, UK, 10–12 September 2018; pp. 1–9.
117
Appl. Sci. 2019, 9, 1449
24. Proakis, J.G.; Manolakis, D.G. Digital Signal Processing, 4th ed.; Pearson: London, UK, 2007.
25. Nakhli Mahal, H.; Yang, K.; Nandi, A.K. Improved Defect Detection Using Adaptive Leaky NLMS Filter in
Guided-Wave Testing of Pipelines. Appl. Sci. 2019, 9, 294. [CrossRef]
27. Gresil, M.; Giurgiutiu, V.; Shen, Y.; Poddar, B. Guidelines for Using the Finite Element Method for Modeling
Guided Lamb Wave Propagation in SHM Processes. In Proceedings of the 6th European Workshop on
Structural Health Monitoring, Dresden, Germany, 3–6 July 2012; pp. 1–8.
28. Miao, H.; Huan, Q.; Wang, Q.; Li, F. Excitation and reception of single torsional wave T(0,1) mode in pipes
using face-shear d24 piezoelectric ring array. Smart Mater. Struct. 2017, 26, 1–9. [CrossRef]
29. Teletestndt. Long Range Guided Wave Testing with Teletest Focus+. 2018. Available online: https://www.
teletestndt.com/ (accessed on 14 January 2019).
30. Dürager, C.; Boller, C.; Cornish, A. Damage feature extraction from measured Lamb wave signals using a
model-based approach. In Proceedings of the 8th European Workshop on SHM (EWSHM), Bilbao, Spain,
5–8 July 2016; Volume 4, pp. 5–8.
118
applied
sciences
Article
A Modified Leakage Localization Method Using
Multilayer Perceptron Neural Networks in a
Pressurized Gas Pipe
Qi Wu 1 and Chang-Myung Lee 1, *
Department of Mechanical and Automotive Engineering, University of Ulsan, 93 Daehak-ro, Nam-Gu,
Ulsan 44610, Korea; wuqidake@gmail.com
* Correspondence: cmlee@ulsan.ac.kr
Received: 15 March 2019; Accepted: 8 May 2019; Published: 13 May 2019
Featured Application: To improve the leakage location accuracy in a pressurized gas pipe, this
paper proposes a modified leakage location method using multilayer perceptron neural networks
based on the acoustic emission signal.
Abstract: Leak detection and location in a gas distribution network are significant issues. The
acoustic emission (AE) technique can be used to locate a pipeline leak. The time delay between
two sensor signals can be determined by the cross-correlation function (CCF), which is a measure
of the similarity of two signals as a function of the time delay between them. Due to the energy
attenuation, dispersion effect and reverberation of the leakage-induced signals in the pipelines,
the CCF location method performs poorly. To improve the leakage location accuracy, this paper
proposes a modified leakage location method based on the AE signal, and combines the modified
generalized cross-correlation location method and the attenuation-based location method using
multilayer perceptron neural networks (MLPNN). In addition, the wave speed was estimated more
accurately by the peak frequency of the leakage-induced AE signal in combination with the group
speed dispersive curve of the fundamental flexural mode. To verify the reliability of the proposed
location method, many tests were performed over a range of leak-sensor distances. The location
results show that compared to using the CCF location method, the MLPNN locator reduces the
average of the relative location errors by 14%, therefore, this proposed method is better than the CCF
method for locating a gas pipe leak.
Keywords: acoustic emission; nondestructive testing; leakage location; fault diagnosis
1. Introduction
Pipelines are used to transport natural gas, water, oil products and other easy-flowing products,
and pipeline leakage is a common phenomenon in the continuous transportation process due to the
corrosion of pipe walls [1], third-party interference [2], aging of the pipes [3], etc. Leakage may lead to
environmental pollution, energy loss, injuries, and even fatalities. Therefore, leakage location is one of
the paramount concerns of pipeline operators and researchers.
Pipeline leakage detection methods include direct and indirect methods [4]. Direct methods
detect the leaked product outside the pipeline using fiber optic sensing, moisture measurements, gas
tracing, vapor sensing, etc. [5]; however, these methods suffer from low sensitivities and expensive
sensors. Indirect methods detect the leakage by measuring the pipeline gas pressure, flow, temperature,
etc. using sensors mounted inside the pipeline; however, many pressurized vessels in industry are
inaccessible, which necessitates nondestructive testing. Acoustic emission (AE) [6] is a nondestructive
testing technique that provides the ability to rapidly locate small-scale leaks [7]; it has thus now become

Appl. Sci. 2019, 9, 1954
a common leak location method. AE signals can be generated by friction, cavitation, impact, or leakage
in a high-pressure pipe. In this study, the AE signals in the pipe wall are detected by AE sensors
mounted on the pipe’s exterior surface.
Leakage location commonly relies on attenuation-based methods or time-of-flight-based
methods [8]. The former methods are based on the reduction in the AE signal strength as the
propagation distance increases. The prerequisite for the latter methods is that the wave speed is a
known constant; however, the wave speed varies as a function of frequency due to the group speed
dispersion nature of gas-leakage-induced guided waves [9]. To determine the speed of the AE signal,
Li et al. [10] derived the wavenumber formulae to predict the wave speeds, and thereby halved the
location error compared to that using the theoretical wave speed. Davoodi and Mostafapour [11]
derived the motion equation of the pipe for the simply supported boundary condition by the standard
form of Donnell’s nonlinear cylindrical shell theory. Nishino et al. [12] generated wide-band cylindrical
waves in aluminum pipes by the laser-ultrasonic method; the experimental results showed good
agreement between the theoretical and experimental group speed dispersion curves. In this study, the
group speed of the leakage-induced signal was estimated by its peak frequency in combination with
the group speed dispersive curve of the fundamental flexural mode.
When leakage-induced signals transmit along the pipe wall, signal distortion arises due to
energy attenuation, dispersion effect [13] and reverberation. The cross-correlation function (CCF) can
determine the time delay based on the similarity of two signals. However, the degree of signal distortion
affects the similarity of two signal waveforms measured by two AE sensors, and this interference
degrades the performance of the CCF method. To improve the leakage location accuracy, Davoodi
and Mostafapour [14,15] located a leak on a high-pressure gas pipe by wavelet transform, filtering
technique, and CCF. Yu et al. [16] located the leakage by dual-tree complex wavelet transform and
singular value decomposition. Brennan et al. [17] estimated the time delay between two channels by the
time and frequency domain methods, with almost identical test results being obtained. Sun et al. [18]
proposed a leakage aperture recognition and location method using the root mean square entropy of
local mean deposition and Wigner-Ville time-frequency analysis. Guo et al. [19] proposed an adaptive
noise cancellation method using the empirical mode decomposition, the test result showed that the
time delay estimation of the leakage-induced signal is more accurate. Yang et al. [20] estimated the
characteristics related to the two propagation channels by blind system identification. The travel
duration of the leakage source signals from the leak point to either sensor was extracted from the
identified propagation channel. This paper proposes a modified generalized cross-correlation (GCC)
location function, which can compensate for the weakening effect of the different propagation paths on
the leakage-induced signals. After the multilayer perceptron neural networks (MLPNN) is trained by
the distances between the impact points and the sensors and the energy ratios of two impulse response
signals measured by two sensors, it can obtain the relationship between the signal energy ratios and
the vibration excitation location, and thus can locate a pipeline leak. This location method can be
used to locate a leak in a pipe made of isotropic materials, such as metallic, plastic, etc., whose elastic
modulus, Poisson’s ratio and density of which should be constant. In this paper, the impulse response
signals are generated by the collision between an impulse hammer and the pipe.
Although the proposed method is applicable for pipes with variable operating pressure, this study
only deals with the location of a single leak source when the gas pressure inside the pipe is constant.
The rest of this paper is organized as follows. Section 2 introduces the modified GCC location method
and wave speed estimation based on modal analysis of gas-leakage-induced guided waves. Section 3
introduces the principle of the leak location method based on the AE signal energy attenuation and
the MLPNN. Section 4 presents the experimental setup, data collection and the signal processing for
leakage location. Section 5 analyzes the experiment results. Section 6 presents the study conclusions.
120
Appl. Sci. 2019, 9, 1954
2. The Principle of Leak Location Method Based on the Generalized Cross-Correlation Function
2.1. The Modified Generalized Cross-Correlation Location Method

Figure 1 presents a pressurized piping system: l is the distance between the two sensors, and l1
and l2 are the distances between the leak point and sensors 1 and 2, respectively.
Figure 1. Schematic of the pressurized piping system.
If there is a leak between the two AE sensors, the two measured signals x1 (t) and x2 (t) can be
mathematically modeled [21] as
x1 (t) = h1 (t) ⊗ s(t) + n1 (t), x2 (t) = h2 (t) ⊗ s(t + d) + n2 (t), (1)
where s(t) is the leakage-induced signal measured at the leak location, the parameter t is the discrete-time
counter, and d denotes the time delay between two measured signals. h1 (t) and h2 (t) are the discrete-time
impulse response functions between the leak point and each sensor, respectively, and they provide the
relationships between an impulse signal and each sensor signal. n1 (t) and n2 (t) are the background
noises picked up by the two sensors, ⊗ is the discrete-time convolution operator. s(t), n1 (t) and n2 (t)
are assumed as stationary random signals. The noises measured by the two sensors are assumed to be
mutually unrelated and unrelated with the leakage-induced signal, so the CCF between two measured
signals with noise and without noise are equivalent, and can be expressed as follow:
Rs1s2 (τ) = Rx1x2 (τ), (2)
where Rs1s2 (τ) and Rx1x2 (τ) are the CCF without noise and with noise, respectively, in terms of the time
delay τ. Since the CCF is sensitive to the non-leak vibration noises [21], the time delay is commonly
estimated by the GCC function [22], which can be written as:

Rx1x2 (τ) = F−1 Ψ g (ω)Sx1x2 (ω) , (3)
where F−1 {·} denotes the inverse discrete Fourier transform, ω denotes the frequency, and Ψg (ω) is the
frequency weighting function of the GCC function. Sx1x2 (ω) is the cross-spectral density between two
measured signals x1 (t) and x2 (t) [23], and can be written as:
∗
X ( ω ) X2 ( ω )
= Sss (ω)H1∗ (ω)H2 (ω)eiωd ,
1
Sx1x2 (ω) = lim E 1 (4)
2π T→∞ T
where E[·] denotes the expectation operator, X1 (ω) and X2 (ω) are the N-point discrete Fourier transforms
(DFT) of x1 (t) and x2 (t), respectively, H1 (ω) and H2 (ω) are the DFT of h1 (t) and h2 (t), respectively, Sss (ω)
is the auto-spectral density of s(t), and (·)* means complex conjugation. Since multiplication in the
frequency domain corresponds to convolution in the time domain [24], the GCC function between
121
Appl. Sci. 2019, 9, 1954
two measured signals in terms of the time delay τ can be derived by substituting Equation (4) into
Equation (3), and expressed as:

Rx1x2 (τ) = F−1 Sss (ω)Ψg (ω)H1∗ (ω)H2 (ω)eiωd = Rll (τ) ⊗ g (τ) ⊗ δ(τ + d), (5)
where Rll (τ) = F−1 {Sss (ω)} is the autocorrelation function of s(t), and Ψg (τ) = F−1 {Ψg (ω)|H1 *(ω)H2 (ω)|},
δ (τ + d) = F−1 {eiωd }. The time delay between two measured signals is equal to the quotient of the
offset of the GCC maximum peak and the sampling frequency. δ (τ + d) is a function equal to zero
everywhere except for τ = −d, and the Dirac Delta function’s integral over the entire real line is equal
to one. If Ψg (ω) = 1, the Dirac delta function δ (τ + d) delayed by d will be broadened by Sss (ω), H1 (ω)
and H2 (ω) [25]. In fact, Sss (ω) is often unknown as it depends on the leakage aperture, the gas pressure
difference between inside and outside the pipe [10], etc. However, H1 (ω) and H2 (ω) can be obtained by
the impulse response method [26], and can be calculated by using:
Yi ( ω )
Hi (ω) = , i = 1, 2, (6)
X (ω)
where the parameter i indicates the sensor serial number, Yi (ω) denotes the DFT of the impulse
response signal measured by the corresponding sensor, and X(ω) is the DFT of the impulse signal
measured by the impact hammer. In this study, the impulse signal is assumed as the Dirac delta
function. Therefore, to compensate for the weakening effect of the pipes on the AE signals, the modified
frequency weighting function of the GCC function is written as:
1
Ψ g (ω) = . (7)
H1∗ (ω)H2 (ω)
The modified GCC function Rx1x2g (τ) can be derived by substituting Equations (4) and (7) into
Equation (3), and expressed as:

Rx1x2 (τ) = F−1 Sss (ω)eiωd = Rll (τ) ⊗ δ(τ + d).
g
(8)
Compared with Equation (5), pre-filter Ψg (ω) removes the effects of the propagation paths on the
AE signal, δ (τ + d) is only broadened by Sss (ω). Therefore, the modified frequency weighting function
can increase the degree of correlation between two measured signals and improve the accuracy of the
time delay estimation. The modified frequency weighting function differs from the PHAT processor
function [25] in that the frequency weighting function involves pre-whitening of the leakage-induced
signal. To reduce the effects of noise on the time delay estimation, only the leakage-induced signal
with greater power spectral density should be extracted and processed by the proposed pre-filter.
Figure 2 shows a schematic of the modified GCC location method, which contains N-point
DFT, complex conjugate, cross power spectrum, pre-filter and N-point inverse discrete Fourier
transform (IDFT).
Figure 2. Schematic of the implementation of the modified GCC location method.
122
Appl. Sci. 2019, 9, 1954
The leakage can be located by:

l + v(ω0 )dest
lest = , (9)
2
where l is the distance between the two sensors, lest is the estimated distance between the leak and one
sensor, and dest is the time delay estimation. The frequency-dependent speed v(ω0 ) [9] of AE signals
along the pipe wall is discussed in Section 2.2. In this study, the relative location error with respect to
the distance between the leak and one sensor is determined as:
|l1 − lest |
δl = × 100%, (10)
l1
The cross-correlation coefficient (CCC) [27] in terms of the time delay τ can be used to evaluate
the correlation of two measured signals and it is given by:
Rx1x2 (τ)
ρ(τ) = , (11)
Rx1x1 (0)Rx2x2 (0)
where Rx1x1 (0) and Rx2x2 (0) are the amplitudes in the autocorrelation functions of x1 (t) and x2 (t),
respectively, for τ = 0. The larger the maximum peak in CCC is, the higher the degree of the correlation
between the two signals is [28].
2.2. Wave Speed Estimation Based on Modal Analysis of Gas-Leakage-Induced Guided Waves
Since the gas solid coupling between the in-pipe gas and the pipe wall can be ignored due to
the acoustic impedance mismatch between them, the gas-filled pipes can be approximated as hollow
cylinders [9]. The dispersive behaviors of the guided waves in a gas pipe can be analyzed by the guided
wave theory of the hollow cylinders [29]. The guided waves in a pipe are grouped into three families:
longitudinal mode, torsional mode, and flexural mode. These modes are denoted by L(0,m), T(0,m)
and F(n,m), respectively, where n and m represent the circumferential and radial modal parameters,
respectively. The theoretical analysis shows that the radial vibration signal collected by AE sensors is
dominated by the dispersive fundamental flexural mode F(1,1) [30].
If the material of the pipe wall is changed, the group speed dispersive curve needs to be calculated
again based on some parameters before determining the group speed of the leak-induced signal, those
parameters are external diameter, thickness, elastic modulus, Poisson’s ratio and density of the pipe
wall. The material and geometric parameters of the given pipe are shown in Table 1. According to
the parameters of the given pipe, the group speed dispersive curves of the flexural modes shown in
Figure 3 can be obtained using the guided wave theory of hollow cylinders in MATLAB R2018b. The
speed of the AE signal can be online estimated by the peak frequency of the AE signal in combination
with the known group speed dispersive curve of the fundamental flexural mode [9].
Table 1. Material and geometric parameters of the given pipe.
a (mm) b (mm) μ ρ (kg/m3 ) G (GPa) E (GPa)

13.9 16.7 0.35 8000 76 205
a is the inner radius of the pipe, b is the outer radius of the pipe, μ is the Poisson’s ratio of the pipe wall material, ρ is
the density of the pipe wall material, G is the shear modulus of the pipe wall material, E is the Young’s modulus of
the pipe wall material.
123
Appl. Sci. 2019, 9, 1954
Figure 3. Group speed dispersive curves of the flexural modes in the frequency band 0–400 kHz for the
given gas pipe.
3. The Principle of the Leak Location Method Based on the AE Signal Energy Attenuation
3.1. AE Signal Energy Attenuation

The energy of the AE signal reduces as the propagation distance increases due to geometric
spreading, structural scattering, and energy absorption [14]. The weakening effect of the pipe on AE
signals is similar with the low-pass filter [10], in that the cut-off frequency decreases as the propagation
distance increases. When the AE signals propagate in one direction along the pipe wall, the energy
attenuation of the AE signal can be described by an exponential decay [31]:
E = E0 e−kli , (12)
where E is the energy of the sensor signal, E0 is the AE signal energy at the leak point and k is the
attenuation coefficient. The AE signal energy is proportional to the integral of the square of the signal
amplitude [32], and can be written as:
ΔT
E= v2 dt, (13)
0
where ΔT represents the length of sample time in seconds, and v represents the instantaneous signal
amplitude in voltages. If the distance between the vibration source and the sensor is constant, the
weakening effect of the pipe on the AE signal will be also constant, irrespective of any leakage-induced
signals or impulse response signals. The closer the sensor is to the leak source, the greater the energy
of the measured signals. Then the energy ratios of two measured signals will gradually increase or
decrease as the distance between the vibration source and one of the two sensors increases. Therefore,
a pipeline leakage can be located by the signal energy ratios without knowing the wave speed or the
time delay estimation.
3.2. Multilayer Perceptron Neural Networks Classifier

The mathematical model of the biological neural network is defined as the artificial neural
network [33]. The relationships between the input and output vectors can be obtained after the artificial
neural networks are trained. Neural network can be applied to solve nonlinear problems, MLPNN is a
commonly used model. Figure 4 shows the perceptron neural networks with three layers. The input
and output layers accept the external input vectors in the learning mode. Every neuron in the input
124
Appl. Sci. 2019, 9, 1954
and hidden layers connects to every neuron in the hidden and output layers, respectively. The input
layer L1 is comprised of P variables x = [x1 , ···, xP ]T , which are processed by the following equation:
P
(1) (1)
aj = ωij xi + ω0 , j = 1, · · · , M, (14)
i=1
where M is the number of neurons in the hidden layer, and the superscript indicates that the parameter
ωij (1) is the connection weight between the neuron xi in the input layer L1 and the neuron zj in the
hidden layer L2 . The parameter ω0 (1) represents the bias of the variable aj , which prevents aj from
becoming zero. Then aj is transformed by an activation function, which is the logistic sigmoid function
in this study, and it is given by:
1
zj = f aj = . (15)
1 + e−a j
Figure 4. The MLPNN with input, hidden, and output layers.
Similarly, the neuron y in the output layer L3 receives the variables zj as shown in the
following equation:

M
(2) (2)
y= ω j z j + ω0 , (16)
j=1
where ωj and ω0 are the connection weight and the bias in the hidden layer, respectively. For the
(2) (2)
regression problem, the aim of the training process is to minimize the squared error function as shown
in the following equation by the gradient descent method:
2
y(ω) − d
E(ω) = , (17)
2
where d is the distance between the vibration source and one sensor, ω is the set of all weights and
biases, and y(ω) represents the neuron in the output layer. All the weights and biases need to be
updated by training the MLPNN on labeled data until the error function becomes small enough or the
maximum iteration is reached; then the MLPNN can be used as a leakage locator.
4. Laboratory Experiments
4.1. Experimental Setup and Data Collection

The experimental setup consists of two parts: a pressurized piping system and a signal acquisition
system, as shown in Figure 5. If the pipe gas pressure is too low, the energy of the leak-induced signal
125
Appl. Sci. 2019, 9, 1954
will be small and difficult to detect. Therefore, in the pressurized piping system, an air compressor
(NH-5, Hanshin product) supplies 7 bar gas pressure to a 15m-long steel pipe (ASTM A 312 TP304)
with a pipe wall thickness and external diameter of 2.8 mm and 33.4 mm, respectively. One end of the
pipe was blocked, and the other end was connected to the air compressor via a hose. To avoid the air
compressor vibration noise affecting the experimental results, the air compressor was placed far away
from the piping system. The pipe was supported by a sponge material to reduce energy dissipation of
the AE signal on the pipe wall to the ground. The diameter of the leakage orifice in the early stage
of a pipeline leakage is usually small. To verify that the proposed method can locate a leak in this
stage, a simulated leak orifice with a diameter of 0.3 mm was punched in the pipe wall. The simulated
leak source, the AE sensor mounted on the pipe wall, an impulse hammer and the placement of two
AE sensors along the pipe are shown in Figures 6–9, respectively. The pipe wall had been cleaned
with sandpaper before the AE sensors were mounted with the magnetic hold downs and vacuum
grease couplant.
Figure 5. Experiment system diagram: (i) a gas pipe, (ii) two AE sensors, (iii) an impulse hammer, (iv)
a data acquisition card, (v) a PC, and (vi) a hose.
Figure 6. A simulated leak orifice of the pipe.
126
Appl. Sci. 2019, 9, 1954
Figure 7. The AE sensor mounted with the magnetic hold down and vacuum grease couplant.
Figure 8. An impulse hammer.
Figure 9. The placement of two AE sensors along the pipe.
The signal acquisition system consists of two R15a sensors with a resonant frequency of 150 kHz,
an impulse hammer (086C03), a data acquisition card (USB-5132, NI product) and a personal computer
(PC) equipped with NI-SCOPE software. Since the sensitivity of the AE sensors is high in the frequency
range between 50 kHz and 400 kHz, data were captured at a sampling rate of 800 kHz in the experiment.
Analog to digital conversion was performed by the data acquisition card connected to the PC. Data
were recorded into the PC for the signal processing and preprocessed by the NI-SCOPE software,
including the preamplification and anti-aliasing filtering.
127
Appl. Sci. 2019, 9, 1954
In the experiment, the training sets for MLPNN include the distances between one sensor and
each impact location and the impulse response signals energy ratios between two sensor signals in
different frequency bands. The impulse signal was measured by the sensor in the impact hammer,
and the impulse response signals were measured by AE sensors when the pipe was tapped by the
impact hammer, and the tapping locations were distributed along a straight line between the two
AE sensors. To avoid the simulated leak orifice affecting the impulse response signal, the impact test
was performed before the simulated leak orifice was punched. The test sets for MLPNN were the
energy ratios of leakage-induced signals in different frequency bands. The leakage-induced signals
were generated by the friction force between the jet of the leaking gas and the simulated leak orifice.
4.2. Signal Processing for Leakage Location

The leakage location procedure for the MLPNN locator is shown in Figure 10, and the MLPNN
locator operates in two different modes:
1. In the learning mode, the training sets of the MLPNN contain the input vectors and the output
vectors. The input vectors are the energy ratios of two impulse response signals in corresponding
frequency bands. Since the impulse response signals are non-stationary signals, and the wavelet packet
analysis is suitable for processing the non-stationary signals, then the impulse response signals were
processed by the wavelet packet transform in this study. The output vectors are the distances between
one sensor and each impact location. The learning procedure for the MLPNN is as follows:
1. Since the energy of a leakage-induced signal at frequency band from 100 to 150 kHz is higher
than at other frequency bands, and three-level decomposition can extract the signal in this
frequency range, so all the impulse response signals are decomposed into three layers of wavelet
packet coefficients, and the signals are reconstructed in eight frequency bands by wavelet packet
coefficients. The Haar wavelet is used in this study.
2. Calculate the energies of the decomposed signals in step 1 by Equation (13); then calculate the
energy ratios of two impulse response signals in the corresponding frequency band.
3. Train the MLPNN with the input and output vectors. The input vectors are the energy ratios in
step 2, and the output vectors are the distances between one sensor and each impact location, and
the epochs in the training method is set to 0.01.
2. In the application mode, the trained MLPNN locator can locate the leakage by using the energy
ratios of the leakage-induced signals in the frequency bands with greater power spectral density. The
leakage location procedures for the MLPNN locator are described as follows:
1. Measure the leakage-induced signals by using the two AE sensors.

2. Determine the peak frequency of the leakage-induced signals by using the power spectral density.
3. Estimate the group speed of the leakage-induced signals by using the peak frequency in step 2
and the known group speed dispersive curve of the fundamental flexural mode.
4. Calculate the CCC by using the CCF method and the modified GCC method.
5. Compare the maximum peak amplitudes of the two CCCs obtained in step 4. ρx1x2 c and ρx1x2 g
are the maximum peak amplitudes of the CCCs obtained by using the CCF method and the
modified GCC method, respectively. If ρx1x2 g is greater than ρx1x2 c , it implies that compared to
the CCC obtained by using CCF method, the CCC obtained by using modified GCC method is
more reliable; then perform step 6, otherwise perform step 7.
6. Calculate the distance between the leakage and sensor 1 by Equation (9). The group speed of
the leakage-induced signal was determined in step 3, and the time delay was determined by the
sampling frequency and the offset of the CCC maximum peak obtained by the modified GCC
method in step 4.
128
Appl. Sci. 2019, 9, 1954
7. Decompose the leakage-induced signals into three layers wavelet packet coefficients, reconstruct
the signals in eight frequency bands by using the wavelet packet coefficients.
8. Calculate the energy ratios of the decomposed signals in the frequency bands with the greater
power spectral density.
9. Locate the leakage by using the energy ratios in step 8 and the trained MLPNN locator.
Figure 10. Flowchart of the leakage location method for the MLPNN locator.
5. Results and Discussion
5.1. Characteristics of the Frequency Domain

Figure 11 shows the power spectral densities of the leakage-induced signals collected at the
distances of 1 m and 2 m away from the simulated leak source. The center frequency of a leak-induced
signal is proportional to the gas jet velocity and inversely proportional to the leakage aperture [10], and
is also related to the resonant frequency of the sensor and the propagation distance between the leak
and the sensor. As shown in Figure 11, the amplitude of the power spectral density decreases with
increasing propagation distance. Since the peak frequency of leakage-induced signals is 124 kHz, the
wave speed can be determined to be 2404 m/s by the peak frequency in combination with the known
group speed dispersive curve of the fundamental flexural mode.
129
Appl. Sci. 2019, 9, 1954
Figure 11. Comparison of the power spectral densities of the leakage-induced signal in the different
distances (L1 = 1 m, 2 m) away from the leak source.
5.2. Characteristics of the Signal Energy Ratios

To verify the consistency of weakening effect of the pipe on two kinds of signals, which are the
impulse response signals and leakage-induced signals measured by maintaining a distance of 14 m
between the two sensors while varying the distance between sensor 1 and the simulated leak orifice.
Since the maximum amount of leakage-induced signal energy occurs in frequencies near 124 kHz, the
signal-to-noise ratio at this frequency band is higher than at other bands, the impulse response signal in
frequencies near 124 kHz was also extracted with wavelet packet transform, and the signal energy ratios
between the two sensors were calculated. Figure 12 shows the fitting curves of the signal energy ratios
as a function of the propagation distance. The horizontal axis denotes the distance between sensor
1 and the simulated leak orifice. The vertical axis denotes the energy ratios of the leakage-induced
signal and the impulse signal measured by the two sensors. The difference between the two fitting
curves is small, which implies that the weakening effect of the pipe on the AE signals is proportional
to the propagation distance, irrespective of any leakage-induced signals or impulse response signals.
Therefore, the signal energy ratio can also be used for locating the vibration excitation source.
Figure 12. Fitting curves of the signal energy ratios of the AE signals measured by the two sensors.
130
Appl. Sci. 2019, 9, 1954
5.3. Characteristics of the Time Domain.

Figure 13a,b show the impulse response signals collected at distances of 0.4 m and 13 m away from
the impact location, respectively. The maximum amplitude of the impulse response signals decreases
as the propagation distance increases, which indicates that the energy of the impulse response signal
decreases due to energy attenuation. Moreover, the closer the vibration source is to the sensor, the
earlier the impulse response signal begins to oscillate.
(a) (b)
Figure 13. Impulse response signals acquired at different distances from sensor 1: (a) 0.4m and (b) 13m.
Figure 14a,b show the CCC obtained by the CCF method and the modified GCC method,
respectively. The time delay estimation only depends on the shift value of the maximum peak in the
cross-correlation coefficient with respect to the central point, and can be calculated as the product of
the sampling period and the shift of the cross-correlation function maximum peak [34]. Two maximum
peaks can be observed, and the time delay estimations are −0.00020 seconds and −0.00034 seconds,
respectively. The distance estimations between the leak and sensor 1 are 1.83 m and 1.99 m, respectively.
Since the relative location error in this study is equal to the quotient of the leakage location error
and the distance between the leak and sensor 1, the real distance between the simulated leak orifice
and sensor 1 is 2.16 m, so the relative location errors are 15% and 8%, respectively. The maximum
peak amplitudes of the two CCCs are 0.168 and 0.236, respectively. Hence, the modified pre-filter can
improve the accuracy of the time delay estimation and increase the degree of the correlation between
the two measured signals.
(a) (b)
Figure 14. Cross-correlation coefficients (CCCs) obtained by two methods: (a) CCF location method,
and (b) modified GCC location method.
131
Appl. Sci. 2019, 9, 1954
5.4. Leakage Location Analysis

To verify the reliability of the MLPNN locator, the gas leakage was located by the CCF location
method and the MLPNN locator under different conditions, which are maintaining a distance of 2.16 m
between the simulated leak orifice and sensor 1 and varying the distance (1.08 m, 2.36 m and 3.54 m)
between the simulated leak orifice and sensor 1. In addition, the group speed of the leakage-induced
signal was determined by the peak frequency of the leakage-induced signal and the group speed
dispersive curve of the fundamental flexural mode. Tables 2 and 3 show the location results obtained
by the CCF method and the MLPNN locator. The relative location errors in Table 3 show that the
relative location errors increase as the distance between the two sensors increases, which implies
that some high frequency components of the leakage-induced signal attenuate as the propagation
distance increases.
Table 2. The location results obtained by the CCF location method.
L (m) 3.24 3.24 3.24 3.24 4.52 4.52 4.52 4.52 5.70 5.70 5.70 5.70
f (kHz) 124.2 123.8 123.4 125.4 105.5 103.4 107.2 106.7 104.7 104.3 104.7 103.5
v (m/s) 2404 2400 2397 2415 2200 2173 2222 2215 2190 2186 2190 2175
l1 (m) 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16
l̂1 (m) 1.78 1.71 1.82 1.72 1.70 1.77 1.75 1.79 1.83 1.74 1.79 1.83
Δl1 (m) 0.38 0.45 0.34 0.44 0.46 0.39 0.41 0.37 0.33 0.42 0.37 0.33
δl1 (%) 17.6 20.1 15.7 20.4 21.3 18.1 19.0 17.1 15.3 19.4 17.1 15.3
L is the distance between the two AE sensors, f is the peak frequency of the leakage-induced signal, v is the group
speed estimation of the leakage-induced signal, l1 is the real distance between the leak and sensor 1, l̂1 is the distance
estimation between the leak and sensor 1, Δl1 is the leakage location error, δl1 is the relative location error.
Table 3. The location results obtained by the MLPNN locator.
L (m) 3.24 3.24 3.24 3.24 4.52 4.52 4.52 4.52 5.70 5.70 5.70 5.70
f (kHz) 124.2 123.8 123.4 125.4 105.5 103.4 107.2 106.7 104.7 104.3 104.7 103.5
v (m/s) 2404 2400 2397 2415 2200 2173 2222 2215 2190 2186 2190 2175
l1 (m) 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16 2.16
l̂1 (m) 2.19 2.18 2.20 2.25 2.26 2.09 2.25 2.27 1.99 2.23 2.28 1.98
Δl1 (m) 0.03 0.02 0.04 0.09 0.10 0.07 0.09 0.11 0.17 0.07 0.12 0.18
δl1 (%) 1.4 0.9 1.9 4.2 4.6 3.2 4.2 5.1 7.9 3.2 5.6 8.3
Figure 15a,b show the histograms of the relative location errors in Tables 2 and 3, respectively. The
vertical axis represents the number of measurements, and the horizontal axis represents the percentage
of relative location error. These two tables show the measurement results of twelve groups by using
the proposed method and the cross-correlation function, respectively, with twelve measurements in
both Figure 15a,b. Despite the small sample size, all the measurement results are typical and can show
the location accuracy of the two location methods. The averages of the relative location errors are 18%
and 4%, respectively, which indicates that the average of the relative location errors obtained by the
MLPNN locator is reduced by 14% compared to that using the CCF location method. Consequently,
the MLPNN locator is more suitable for locating a leak in a gas pipe than the CCF location method
is. However, the MLPNN locator suffers the drawback of needing to be trained again if one of the
following experimental conditions changes: for example, the distance between the two AE sensors, the
geometric parameters of the pipe wall, or the material of the pipe wall.
132
Appl. Sci. 2019, 9, 1954
(a) (b)
Figure 15. Histograms of the relative location errors obtained by two methods: (a) CCF location method
and (b) MLPNN locator.
6. Conclusions
This paper has proposed a leakage location method using the modified generalized
cross-correlation (GCC) location method in combination with the attenuation-based location method
using multilayer perceptron neural networks (MLPNN). The GCC location method can compensate for
the weakening effect of the different propagation paths on the leakage-induced signals, can increase
the degree of the correlation between two measured signals and improve the accuracy of the time
delay estimation. Unlike the cross-correlation function (CCF) location method, the MLPNN locator
locates a leak not only by the time delay estimation between two measured signals but also by the
energy ratios of two leakage-induced signals measured by two sensors in the frequency bands with
greater power spectral density. In addition, the wave speed was determined more accurately by using
the peak frequency in combination with the group speed dispersive curve of the fundamental flexural
mode. The MLPNN locator was experimentally verified by using a pressurized piping system and a
signal acquisition system. The average of the relative location errors obtained by the MLPNN locator
was reduced by 14% compared to that using the CCF location method. Hence, the MLPNN locator is
more suitable for locating a leak in a gas pipe than the CCF location method is.
Data Availability: The data used to support the findings of this study are available from the corresponding author
upon request.
Author Contributions: Conceptualization, C.-M.L.; methodology, Q.W. and C.-M.L.; software, Q.W.; validation,
Q.W.; formal analysis, Q.W.; investigation, Q.W.; resources, C.-M.L.; data curation, X.X.; writing—original
draft preparation, Q.W.; writing—review and editing, Q.W.; visualization, Q.W.; supervision, C.-M.L.; project
administration, C.-M.L.; funding acquisition, C.-M.L.
Funding: This research was funded by a 2017 grant from the Russian Science Foundation (Project No. 17-19-01389).
References
1. Edalati, K.; Rastkhah, N.; Kermani, A.; Seiedi, M.; Movafeghi, A. The use of radiography for thickness
measurement and corrosion monitoring in pipes. Int. J. Press. Vessels Pip. 2006, 83, 736–741. [CrossRef]
2. Liang, W.; Hu, J.; Zhang, L.; Guo, C.; Lin, W. Assessing and classifying risk of pipeline third-party interference
based on fault tree and SOM. Eng. Appl. Artif. Intell. 2012, 25, 594–608. [CrossRef]
3. Vargas-Arista, B.; Hallen, J.M.; Albiter, A. Effect of artificial aging on the microstructure of weldment on API
5L X-52 steel pipe. Mater. Charact. 2007, 58, 721–729. [CrossRef]
4. Ferraz, I.N.; Garcia, A.C.; Bernardini, F.C. Artificial Neural Networks Ensemble Used for Pipeline Leak
Detection Systems. In Proceedings of the 7th International Pipeline Conference IPC 2008, Calgary, AB,
Canada, 29 September–3 October 2008; pp. 739–747.
133
Appl. Sci. 2019, 9, 1954
5. Nehorai, A.; Paldi, E.; Porat, B. Detection and Localization of Vapor-Emitting Sources. IEEE Trans. Signal
Process. 1995, 43, 243–253. [CrossRef]
6. Xiao, Q.; Li, J.; Bai, Z.; Sun, J.; Zhou, N.; Zeng, Z.A. Small Leak Detection Method Based on VMD Adaptive
De-Noising and Ambiguity Correlation Classification Intended for Natural Gas Pipelines. Sensors 2016, 16,
2116. [CrossRef] [PubMed]
7. Faerman, V.A.; Cheremnov, A.G.; Avramchuk, V.V.; Luneva, E.E. Prospects of Frequency-Time Correlation
Analysis for Detecting Pipeline Leaks by Acoustic Emission Method. IOP Conf. Ser. Earth Environ. Sci. 2014,
21, 012041. [CrossRef]
8. Rewerts, L.E.; Roberts, R.R.; Clark, M.A. Dispersion compensation in acoustic emission pipeline leak location.
In Review of Progress in Quantitative Nondestructive Evaluation; Springer: Boston, MA, USA, 1997; Volume 16,
pp. 427–434.
9. Li, S.; Wen, Y.; Li, P.; Yang, J.; Dong, X.; Mu, Y. Leak location in gas pipelines using cross-time-frequency
spectrum of leakage-induced acoustic vibrations. J. Sound Vib. 2014, 333, 3889–3903. [CrossRef]
10. Li, S.; Wen, Y.; Li, P.; Yang, J.; Yang, L. Determination of acoustic speed for improving leak detection and
location in gas pipelines. Am. Inst. Phys. 2014, 85, 024901. [CrossRef]
11. Davoodi, S.; Mostafapour, A. Modeling Acoustic Emission Signals Caused by Leakage in Pressurized Gas
Pipe. J. Nondestruct. Eval. 2013, 32, 67–80. [CrossRef]
12. Nishino, H.; Takashina, S.; Uchida, F.; Takemoto, M.; Ono, K. Modal analysis of hollow cylindrical guided
waves and applications. Jpn. J. Appl. Phys. 2001, 40, 364–370. [CrossRef]
13. Grabec, I. Application of correlation techniques for localization of acoustic emission sources. Ultrasonics
1978, 16, 111–115. [CrossRef]
14. Davoodi, S.; Mostafapour, A. Gas leak locating in steel pipe using wavelet transform and cross-correlation
method. Int. J. Adv. Manuf. Technol. 2013, 70, 1125–1135. [CrossRef]
15. Mostafapour, A.; Davoodi, S. Leakage Locating in Underground High Pressure Gas Pipe by Acoustic
Emission Method. J. Nondestruct. Eval. 2012, 32, 113–123. [CrossRef]
16. Yu, X.C.; Liang, W.; Zhang, L.B.; Jin, H.; Qiu, J.W. Dual-tree complex wavelet transform and SVD based
acoustic noise reduction and its application in leak detection for natural gas pipeline. Mech. Syst. Signal
Process. 2016, 72–73, 266–285. [CrossRef]
17. Brennan, M.J.; Gao, Y.; Joseph, P.F. On the relationship between time and frequency domain methods in time
delay estimation for leak detection in water distribution pipes. J. Sound Vib. 2007, 304, 213–223. [CrossRef]
18. Sun, J.D.; Xiao, Q.Y.; Wen, J.T.; Zhang, Y. Natural gas pipeline leak aperture identification and location based
on local mean decomposition analysis. Measurement 2016, 79, 147–157. [CrossRef]
19. Guo, C.; Wen, Y.; Li, P.; Wen, J. Adaptive noise cancellation based on EMD in water-supply pipeline leak
detection. Measurement 2016, 79, 188–197. [CrossRef]
20. Yang, J.; Wen, Y.; Li, P. Leak location using blind system identification in water distribution pipelines. J. Sound
Vib. 2008, 310, 134–148. [CrossRef]
21. Choi, J.; Shin, J.; Song, C.; Han, S.; Park, D.I. Leak Detection and Location of Water Pipes Using Vibration
Sensors and Modified ML Prefilter. Sensors 2017, 17, 2104. [CrossRef]
22. Knapp, C.; Carter, G. The generalized correlation method for estimation of time delay. IEEE Trans. Accoust.
Speech Signal Process. 1976, 244, 320–327. [CrossRef]
23. Gao, Y.; Brennan, M.J.; Joseph, P.F.; Muggleton, J.M.; Hunaidi, O. A model of the correlation function of leak
noise in buried plastic pipes. J. Sound Vib. 2004, 277, 133–148. [CrossRef]
24. Gao, Y.; Brennan, M.J.; Joseph, P.F.; Muggleton, J.M.; Hunaidi, O. On the selection of acoustic/vibration
sensors for leak detection in plastic water pipes. J. Sound Vib. 2005, 283, 927–941. [CrossRef]
25. Gao, Y.; Brennan, M.J.; Joseph, P.F. A comparison of time delay estimators for the detection of leak noise
signals in plastic water distribution pipes. J. Sound Vib. 2006, 292, 552–570. [CrossRef]
26. Hur, J.; Kim, S.; Kim, H. Water hammer analysis that uses the impulse response method for a reservoir-pump
pipeline system. J. Mech. Sci. Technol. 2017, 31, 4833–4840. [CrossRef]
27. Almeida, F.C.; Brennan, M.J.; Joseph, P.F.; Dray, S.; Whitfield, S.; Paschoalini, A.T. Towards an in-situ
measurement of wave velocity in buried plastic water distribution pipes for the purposes of leak location.
J. Sound Vib. 2015, 359, 40–55. [CrossRef]
134
Appl. Sci. 2019, 9, 1954
28. Li, S.; Zhang, J.; Yan, D.; Wang, P.; Huang, Q.; Zhao, X.; Cheng, Y.; Zhou, Q.; Xiang, N.; Dong, T. Leak
detection and location in gas pipelines by extraction of cross spectrum of single non-dispersive guided wave
modes. J. Loss Prev. Process Ind. 2016, 44, 255–262. [CrossRef]
29. Xiao, Q.; Li, J.; Sun, J.; Feng, H.; Jin, S. Natural-gas pipeline leak location using variational mode decomposition
analysis and cross-time–frequency spectrum. Measurement 2018, 124, 163–172. [CrossRef]
30. Li, S.; Wen, Y.; Li, P.; Yang, J.; Wen, J. Modal analysis of leakage-induced acoustic vibrations in different
directions for leak detection and location in fluid-filled pipelines. In Proceedings of the IEEE International
Ultrasonics Symposium, Chicago, IL, USA, 3–6 September 2014.
31. Reuben, R.L.; Steel, J.A.; Shehadeh, M. Acoustic Emission Source Location for Steel Pipe and Pipeline
Applications: The Role of Arrival Time Estimation. Proc. Inst. Mech. Eng. Part E J. Process Mech. Eng. 2006,
220, 121–133.
32. Mostafapour, A.; Davoodi, S. Continuous leakage location in noisy environment using modal and wavelet
analysis with one AE sensor. Ultrasonics 2015, 62, 305–311. [CrossRef]
33. Gopi, E.S. Algorithm Collections for Digital Signal Processing Applications Using Matlab; Springer: Dordrecht,
The Netherlands, 2010.
34. Ionel, R.; Ionel, S.; Bauer, P.; Quint, F. Water leakage monitoring education: cross correlation study via
spectral whitening. In Proceedings of the IECON 2014—40th Annual Conference of the IEEE Industrial
Electronics Society, Dallas, TX, USA, 29 October–1 November 2014.
135
applied
sciences
Article
Using the Partial Wave Method for Wave Structure
Calculation and the Conceptual Interpretation of
Elastodynamic Guided Waves
Christopher Hakoda and Cliff J. Lissenden *
Department of Engineering Science and Mechanics, Pennsylvania State University,
University Park, PA 16802, USA; cnh137@psu.edu
* Correspondence: Lissenden@psu.edu
Received: 8 May 2018; Accepted: 8 June 2018; Published: 12 June 2018
Featured Application: Shortcut to wave-structure calculation for elastodynamic guided waves

and prediction of acoustic leakage behavior.
Abstract: The partial-wave method takes advantage of the Christoffel equation’s generality to represent
waves within a waveguide. More specifically, the partial-wave method is well known for its usefulness
when calculating dispersion curves for multilayered and/or anisotropic plates. That is, it is a vital
component of the transfer-matrix method and the global-matrix method, which are used for dispersion
curve calculation. The literature suggests that the method is also exceptionally useful for conceptual
interpretation, but gives very few examples or instruction on how this can be done. In this paper,
we expand on this topic of conceptual interpretation by addressing Rayleigh waves, Stoneley waves,
shear horizontal waves, and Lamb waves. We demonstrate that all of these guided waves can be described
using the partial-wave method, which establishes a common foundation on which many elastodynamic
guided waves can be compared, translated, and interpreted. For Lamb waves specifically, we identify
the characteristics of guided wave modes that have not been formally discussed in the literature.
Additionally, we use what is demonstrated in the body of the paper to investigate the leaky characteristics
of Lamb waves, which eventually leads to finding a correlation between oblique bulk wave propagation
in the waveguide and the transmission amplitude ratios found in the literature.
Keywords: partial wave method; slowness curves; lamb wave; stoneley wave; mode sorting;
acoustic leakage; rayleigh wave; surface waves; elastodynamics
1. Introduction
In this paper, we demonstrate how the partial-wave method, as first introduced by Solie and Auld [1],
can be used to gain a new perspective on guided waves. This new perspective adds to our understanding
of the structure of guided waves, and provides valuable insight into the leaky characteristics of Lamb waves
(which are not to be confused with the study of leaky Lamb waves and their corresponding dispersion
curve solutions).
At present, our understanding of guided waves is primarily informed by the derivation of
characteristic equations and dispersion curve solutions. The dispersion curve solutions in particular
have become significantly more accessible with the widespread use of the semi-analytical finite element
(SAFE) method. However, the SAFE method does not provide the same level of understanding that
would be possible by analyzing a characteristic equation. The most well-known example of this is that the
two-part Lamb wave characteristic equation shows that all of the Lamb waves can be divided into two types:
symmetric and anti-symmetric. Hence, the importance of the present work. However, this paper will not

Appl. Sci. 2018, 8, 966
cover the calculation of dispersion curves, attenuation, energy loss, or anisotropic materials. Neither will it
cover the derivation of characteristic equations for various guided waves.
For the purpose of clarity, in this paper, bulk waves are plane waves that travel through the bulk of
a media, and surface waves are plane waves that travel along the surface of a media. It is emphasized that
all of the waves are considered to have a planar wavefront.
Sections 2 and 3 will provide a brief history and the fundamentals of the partial-wave method and
the Christoffel equation, respectively. These sections give special attention to the diction used and the
assumptions made, as they are pertinent to how the interpretation presented in this paper deviates from
the literature. Section 4 describes this new method of interpretation. It discusses in detail how the slowness
curve solutions of the Christoffel equation can be used to interpret the characteristics of most guided
waves, and the Lamb wave solutions in particular. Sections 5 and 6 apply the new method of interpretation
to a variety of guided waves, which include Rayleigh waves, Stoneley waves, Lamb waves, and shear
horizontal waves.
Section 7 focuses on Lamb waves, and a method developed in Section 6 is used to investigate the
contributions of each partial wave to Lamb waves from 50 kHz to 20 MHz for a 1-mm thick aluminum plate.
This section also identifies characteristics of Lamb waves that have not been formally discussed in
the literature.
Section 8 uses the results from the previous sections to develop an understanding of the leaky
characteristics of Lamb waves. Various characteristics of the acoustic leakage from the waveguide can
be easily determined, including the angle of propagation and the amplitude of emission based on the
interface conditions. A correlation between the amount of acoustic leakage and the refraction qualities of
obliquely travelling bulk waves in the guided wave is also identified.
2. The Partial Wave Method

The term ‘partial-wave method’ first appeared in a paper by Solie and Auld in 1973 [1]; specifically,
it was called “the method of partial waves (or transverse resonance)”. This method was described as
follows: “In this method, the plate wave solutions are constructed from simple exponential-type waves
(partial waves), which reflect back and forth between the boundaries of the plates . . . ” They go on to
write, “A basic principle of the partial wave method is that the partial waves are coupled to each other by
reflections at the plate boundaries”.
In Auld’s 1990 monograph [2], he discusses the method further: “[the superposition of partial waves
method] can be further simplified by making use of the transverse resonance principle, which has been used
to great advantage in electromagnetism” [emphasis added by Auld]. That is, what is known colloquially
as the partial-wave method can be broken down into two parts. First, there is the superposition of the
partial-waves method, which uses the principle of superposition with the Christoffel equation’s solutions
(i.e., partial waves) [2]. Second, there is the transverse resonance principle, which assumes that the
standing wave perpendicular to wave propagation can only exist if the partial waves satisfy the relevant
boundary conditions [2,3]. Modern application of the partial-wave method includes the calculation of
dispersion curves using the global-matrix method or the transfer-matrix method for anisotropic plates
and/or multilayered plates [4,5].
3. The Christoffel Equation and the Use of Slowness Curves

The Christoffel equation is critical to the flexibility and strength of the partial-wave method.
The Christoffel equation can be derived by assuming a plane wave (Equation (4)) propagating in a solid
elastic media (Equations (1)–(3)).
σij,j = ρui,tt (1)
σij = Cijkl kl (2)

1
ij = u + u j,i (3)
2 i,j
138
Appl. Sci. 2018, 8, 966
ui ( x, y, z, t) = Ui ei(−k j r j +ωt) (4)
Here σij , ij , and ui are the stress, strain, and displacement; ρ and Cijkl are the mass density and elastic
stiffness; k j and ω are the wave vector and angular frequency; i, j, k, l = x, y, z, and t is the time variable.
When combined, Equations (1)–(4) yield the Christoffel equation:
Cijkl k k Ul k j − ρω 2 Ui = 0 (5)
When Equation (5) is solved, a slowness curve/surface can be calculated and plotted in an inverse
phase velocity space.
3.1. Deriving the Expression for the Slowness Curves of an Isotropic Solid
For an isotropic linear elastic solid, Equation (2) can be replaced with:
σij = λkk δij + 2μij (6)
where λ and μ are the Lamé parameters and δ is the Kronecker delta. As a side note, in addition to
replacing Equation (2) with Equation (6), if Equation (4) is replaced with:
uiL ( x, t) = Uo x̂ei(−k x x+ωt) (7)
or:
uiS ( x, t) = Uo ẑei(−k x x+ωt) (8)
then, the Christoffel equation can be used to derive the relationship between the Lamé parameters,
the longitudinal wave speed, cL , and the shear wave speed, cS :
λ + 2μ = ρc2L (9)
μ = ρc2S (10)
where x̂ and ẑ are the unit vectors along the x-axis and the z-axis. Using Equation (6), assuming ky = kz = 0
(i.e., wave propagation in the x–z plane), and dividing through by ω2 , Equation (5) can be rearranged into
matrix form:
Aui = 0 (11)
where:
⎡ ⎤
(λ + 2μ)c− −2
x + μcz − ρ
2 0 c− 1 −1
x cz (λ + μ)
⎢ ⎥
A=⎣ 0 μ c−
x + cz
2 −2 − ρ 0 ⎦ (12)
c− 1 −1
x cz (λ + μ) 0 μc−
x
2 + ( λ + 2μ ) c−2 − ρ
z
Equations (11) and (12) represent a system of linear homogeneous equations, where c−
x = k x /ω
1
and cz−1 = kz /ω. The non-trivial solution to Equation (11) only exists if:
det( A) = 0 (13)
Expanding Equation (13) based on Equation (12) results in:

2
ρ − (λ + 2μ) c− 2 −2
x + cz ρ − μ cz−2 + c−
x
2
=0 (14)
The solutions to the characteristic equation (Equation (14)) are:

ρ
c− −2
x + cz =
2
(15)
(λ + 2μ)
139
Appl. Sci. 2018, 8, 966
ρ
c− −2
x + cz =
2
(16)
μ
If Equations (9) and (10) are considered, the right-hand side of Equations (15) and (16) become
c− 2 −2
L and cS , respectively. These represent the slowness curve solutions to the Christoffel equation of
an isotropic material. It should be noted that in Equation (14), the squared term represents the multiplicity
in the shear wave’s slowness curve solutions.
3.2. Interpretation of Slowness Curve Solutions

From the previous section, it is clear that the solutions to the Christoffel equation do not take into
consideration the boundary conditions (BCs) of the problem. That is, the solution only depends on the
material properties of the elastic solid in which the plane waves propagate. As a result, in the literature, it is
shown that the Christoffel equation can be used to determine the reflection and refraction characteristics
of plane waves at oblique incidence to an interface between anisotropic elastic solids [6–8]. Additionally,
the energy velocity vector and the bulk wave’s skew angle can be calculated [4].
Slowness curves are often plotted in terms of the ‘inverse phase velocity’. However, since phase
velocity is not a vector, the use of the term ‘inverse phase velocity’ to describe the slowness vector can
be confusing. A different notation that may be easier to conceptually understand is plotting slowness
curves on the kz /ω and k x /ω axis instead of the c− 1 −1
z versus cx axis. Therefore, if a ray is drawn from the
origin to a point on the slowness curve, it is understood that its unit vector describes the wave propagation
direction in the x–z plane.
Since the focus of the partial-wave method is on guided waves, for the purposes of this paper,
it is assumed that the wave is propagating parallel to the boundary/interface, which we take to be in
the x-direction. The z-axis is through the thickness of the elastic solid. This coordinate system will
be maintained throughout the entirety of the paper, with additional details depending on the problem
being considered. Further, continuing to assume that ky = 0, and using the frequency–wavenumber (ω-k x )
pairs obtained from any one of the many methods for calculating dispersion curves (i.e., enforcing the
BC and solving the eigenvalue problem), the Christoffel equation becomes a complex valued quadratic
eigenvalue problem in three dimensions. The solution, should it exist, will always be composed of
six complex valued eigensolutions. These six solutions are the partial waves of the partial-wave method.
The partial waves can be paired up into three pairs, and each pair is composed of an upward and
a downward propagating wave (i.e., the magnitudes of the eigenvalues are the same, but the signs
are opposite) [1,4].
In Auld [2], there is a discussion of regions for an increasing k x , in which the partial waves have
specific characteristics. Also, in Auld [2], it is recognized that at an interface, k x is the same for all of the
partial waves in order to satisfy Snell’s law, and that a given guided wave must theoretically have a single
k x value for a given ω. Auld gives an example for an isotropic material, which has two slowness curves
and three regions (see Equations (15) and (16)):
• Region 1: 0 < c p < cS All six partial waves are surface waves (imaginary-valued kz ) (see Figure 1a).
• Region 2: cS < c p < c L : Two partial waves are surface waves (imaginary-valued kz ), and four are
bulk waves (real-valued kz ) (see Figure 1b).
• Region 3: c L < c p : All six partial waves are bulk waves (real-valued kz ) (see Figure 1c).
The inverses of the shear and longitudinal group velocities (c− 1 −1

S and c L ) are also the radii of the
two slowness curves, which are circles for an isotropic material. By Crandall’s paper [6], it is also understood
that outside of these circles, the value for k Z becomes imaginary, and the magnitude traces a hyperbola.
Auld also gives examples of slowness curve representations for a shear horizontal (SH) mode
composed of obliquely travelling bulk waves, and a piezoelectric plate with leaky electromagnetic
radiation [2].
This concludes the state-of-the-art section of the paper. How these theories and concepts allow us to
reinterpret our understanding of guided waves in a novel way is discussed from here onward.
140
Appl. Sci. 2018, 8, 966
Figure 1. Depiction of the characteristics of (a) region 1, (b) region 2, and (c) region 3 with respect to the
slowness curves of an isotropic solid. Slowness curves with solid lines denote real-valued kz solutions,
and dotted lines denote imaginary-valued kz solutions. This figure is a compilation of elements from the
figures in Auld [2] and Crandall [6].
4. Reinterpreting the Dispersion Curves

As mentioned previously, the Christoffel equation is only dependent upon the material properties
of the elastic solid in which the plane waves propagate. The BCs are not used until the partial-wave
method is applied. Not only that, but the boundary conditions are not limited to the conventional
traction-free BCs (e.g., Lamb waves). As a result, the partial-wave method is not limited to any specific
type of waveguide. In fact, most types of guided waves should be representable using slowness curves.
For some, such as the shear horizontal guided waves, this approach is more cumbersome than the
standard derivation. However, its value lies in how it provides a fundamental foundation for all of the
guided waves, and a common method for interpreting and comparing each one. When considering
reflection/transmission, mode conversion, or multi-mode propagation, this perspective is quite beneficial.
In modern literature, perhaps the most commonly used dispersion curves are of Lamb waves
in an aluminum plate in terms of phase velocity versus the frequency–thickness product. Using the
three regions depicted in Figure 1, we can divide the phase velocity dispersion curves into three regions,
as shown in Figure 2. By doing this, it is possible to view how the dispersion curve solutions match up with
the partial waves’ characteristics. For example, this is a good way to dispel the potential misconception
that guided waves are composed of multiple reflecting bulk waves. That is, the original assumption of
plane wave propagation (Equation (4)), which is fundamental to the Christoffel equation, does not exclude
the surface wave solution (i.e., imaginary-valued kz ) as k x increases. It would be more accurate to say
that guided waves are a superposition of reflecting bulk waves and surface waves with elliptical particle
motion (this will be shown in the later calculations and examples).
In fact, guided waves composed solely of reflecting bulk waves exist only in Region 3 (c), where they
are highly dispersive. However, there is one exception to this rule: there exists a range of phase velocities
in Region 2 (Figure 1b), where only bulk waves propagate, which will be discussed in a later section.
The A0 mode and Rayleigh waves exist in Region 1 (Figure 1a), where each guided wave can be
represented in terms of six surface waves. The low-frequency S0 mode is in Region 2 (b), and is composed
of two surface waves and two bulk waves (as will be shown later in the paper), excluding the two shear
horizontal partial waves.
The line separating Region 1 (Figure 1a) and Region 2 (Figure 1b) has great significance, since
many Lamb wave modes appear nearly non-dispersive near it. That is, the higher-order Lamb waves
approach the shear wave speed, and the A0 and S0 modes approach the Rayleigh wave speed for large
ω values. The shear horizontal modes also converge to this line for large ω values. The line separating
Region 2 (Figure 1b) and Region 3 (Figure 1c) has many higher-order Lamb modes, which appear nearly
non-dispersive close to it as well.
141
Appl. Sci. 2018, 8, 966
Figure 2. Lamb wave dispersion curves for an aluminum plate divided according to the three regions
shown in Figure 1.
The A0 and S0 Lamb wave modes’ tendency to approach the Rayleigh wave phase velocity can
also be explained using this reinterpretation, since increasing k x will increase the magnitude of the
imaginary-valued kz (see Figure 1a). Once kz increases sufficiently, the surface waves decay with the
depth from the surface before interacting with the opposite traction-free boundary, thereby imitating
a Rayleigh wave. This also explains why quasi-Rayleigh waves [9] have similar wave structures to
Rayleigh waves. That is, the quasi-Rayleigh wave and the Rayleigh wave are both composed of partial
waves (with slight differences in amplitude and exponential decay).
Many higher-order Lamb modes have wave structures that are sinusoidal through the thickness,
but have slightly larger or smaller amplitudes near the traction-free boundaries. In Region 2 (Figure 1b),
this could be explained by surface waves affecting the wave structure near the traction-free boundaries,
while the reflecting bulk waves create the sinusoidal appearance.
This reinterpretation of guided waves provides for the relatively simple calculation of wave structures
based on the superposition of the partial-wave eigensolutions. Based on the preceding concepts and with
wave propagation in the x-direction, c p = ω/k x = cx . Solving Equations (15) and (16) for c− 1
z results in:

c−
z =±
1
c− −2
L − cx
2
(17)
and:
c−
z =±
1
cS−2 − c−
x
2
(18)
respectively. Equations (9), (10) and (17) can then be substituted into Equation (11) to get matrix A as
−1
a function of c−1
x and c L :
⎡ ⎤
(λ + μ) c− 2 −2
L − cx ± c− −2 −1
L − c x c x (λ + μ)
0 2
⎢ ⎥
A( L) := ⎢
⎣ 0 −(λ + μ)c−
L
2
0 ⎦ (19)

± c− −2 −1
L − c x c x (λ + μ)
2 0 (λ + μ) −c−
x
2
And Equations (9), (10) and (18) can be substituted into Equation (11) to get matrix A as a function of
−1
and c−1
x and cS :
⎡ ⎤
−2 −2 − μc−2
(λ + 2μ)c−
x + μ cS − c x
2
S 0 ± c− −2 −1
S − c x c x (λ + μ)
2
⎢ ⎥
A(S) : = ⎢
⎣ 0 0 0
⎥
⎦ (20)
− −2 − −2
± c−
S
2
− c− 2 −1
x c x (λ + μ)
0 μc x + (λ + 2μ) cS − c x − μcS
2 2
142
Appl. Sci. 2018, 8, 966
Finding the reduced row echelon form of A(L) and A(S) gives the solution to the eigenvectors,
which contain, at most, two free variables (i.e., m and p). Considering A(L) and A(S) , and also the “±”
terms in them, which correspond with the sign of kz , results in four possible eigenvectors (not including
the multiplicity). For the positive kz eigenvalue of the longitudinal wave, the eigenvector is:
⎡ −1 ⎤
cx m
(+ L) ⎢ c− 2 −2
L −c x ⎥
ui =⎢
⎣
⎥
⎦ (21)
0
m
While for the negative k z eigenvalue of the longitudinal wave, the eigenvector is:
⎡ c− 1 ⎤
− x
m
(− L) ⎢ c− 2 −2
L −c x ⎥
ui =⎢
⎣
⎥
⎦ (22)
0
m
For the positive k z eigenvalue of the shear wave, the eigenvector is:
⎡ ⎤
− cS−2 −c−
x
2
⎢ m ⎥
(+S) c− 1
ui =⎢
⎣
x ⎥
⎦ (23)
p
m
While for the shear wave’s negative k z eigenvalue, the eigenvector is:
⎡ ⎤
c− 2 −2
S −c x
⎢ m ⎥
(−S) c− 1
ui =⎢
⎣
x ⎥
⎦ (24)
p
m
The superscripts L and S correspond to the longitudinal and the shear waves, respectively. The S can
be made more specific by using SV and SH to denote the shear vertical and shear horizontal polarizations.
Examining Equations (21) and (22), it is clear that for c− 1 > c−1 , the square root becomes
x L
imaginary-valued, and for c− 1 < c−1 , the square-root becomes real-valued. These two scenarios correspond
x L
with the surface wave and bulk wave solutions, respectively. The same holds true for Equations (23) and
−1 −1
(24) with cS instead of cL , with p = 0. For c− −1
x c L or cx cS and positive kz , the corresponding
1
eigenvector is: ⎡ ⎤
−im
(+) ⎢ ⎥
ui = ⎣ 0 ⎦ (25)
m
and for negative k z it is: ⎡ ⎤
im
(−) ⎢ ⎥
ui =⎣ 0 ⎦ (26)
m
Hence, for very large c− 1

x values (relative to their respective wave speeds), the eigenvectors become
circularly polarized, regardless of whether it is a shear wave or longitudinal wave. Lastly, for a given k x
and ω, the kz values of the four partial waves (not including the multiplicity of the shear waves) can be
calculated using Equations (15) and (16):

(± L)
kz = ±ω c− −2
L − cx
2
(27)
143
Appl. Sci. 2018, 8, 966

(±S)
kz = ±ω c− −2
S − cx
2
(28)
As implied previously, since the solution of the Christoffel equation does not depend on the
boundary/interface conditions, a linear combination of the partial waves should be able to represent
most elastodynamic guided waves, regardless of the boundary/interface conditions. We demonstrate this
in the following sections. The material properties shown in Table 1 will be used. For each partial-wave
eigenvector calculated in the following examples, the free variable is set equal to unity; then, the resulting
eigenvector is normalized into a unit vector by dividing by its magnitude.
Table 1. Material properties used for the calculations shown in this paper.
Aluminum [4] Tungsten [4] Air [10]

ρB , kg/m3 2700 19,250 1.2
λB , GPa 56.98 199.4 -
μ B , GPa 25.95 158.6 -
cL , m/s 6350 5180 343
cS , m/s 3100 2870 -
5. Rayleigh Waves and Generalized Rayleigh Waves
5.1. Rayleigh Waves

The Rayleigh wave, which travels at the surface of a half-space, is a good starting example, since its
solution is well-known, and this different approach will mirror the conventional derivation in the literature.
The term ‘generalized Rayleigh waves’ will be used to refer to a class of wave propagation solutions
developed following the original Rayleigh wave derivation, as was done by Stoneley [11]. This will be
considered later in section V.B.
Assuming an aluminum half-space with the boundary at z = 0, an example Rayleigh wave calculation
is completed using the previously mentioned methods. The eigensolutions for the six partial waves at
a frequency of 3.188 × 106 rad/s and a wavenumber of 1103.4 rad/m are shown in Table 2. The frequency
and wavenumber were chosen to match the non-dispersive phase velocity solution to the standard Rayleigh
wave problem.
Table 2. Eigenvalues and normalized eigenvectors from the Christoffel equation for aluminum at
ω = 3.188 × 106 rad/s and k x = 1103.4 rad/m.
Partial Waves kz , rad/m ux uy uz

(+L) 982.6i −0.747i 0 0.665
(−L) −982.6i 0.747i 0 0.665
(+SV) 400.0i −0.341i 0 0.940
(−SV) −400.0i 0.341i 0 0.940
(+SH) 400.0i 0 1 0
(−SH) −400.0i 0 −1 0
There are several notable results in Table 2:
• All of the k z values are imaginary, since Rayleigh waves exist in Region 1 (Figure 1a);
(± L) (±SV ) (±SH )
• kz is larger in magnitude than k z and k z because of the hyperbolic shape of the
imaginary-valued slowness curves;
(+SV ) (−SV ) (+SH ) (−SH )
• kz , kz , kz , and k z have a multiplicity due to the isotropic material properties;
• The eigenvectors are complex, which suggests elliptical particle motion [9];
(+SH ) (−SH )
• ui and ui are not complex-valued.
As shown in Figure 3, only the exponentially decaying solutions are necessary for the Rayleigh
wave solution, because the domain is a half-space.
144
Appl. Sci. 2018, 8, 966
Figure 3. Conceptual depiction of the partial waves, which are necessary for the Rayleigh-wave
problem, overlaid on a traction-free boundary of an aluminum half-space.
Although the derivation method is different, closer inspection of the k z solutions shows that they
match the values of q and s given by Viktorov [9]. In the conventional derivation, the Rayleigh wave
solution is a superposition of two exponentially decaying terms that are dependent upon the phase
velocity and material properties, so this close agreement is expected. A superposition of the two partial
waves, when ui ( x, z, t) = ui (z)ei(−k x x+ωt) , can be represented as:

(− L) −ik(− L) (−SV ) −ik(− SV )
ui (z) = B(− L) ui e z z + B(−SV ) ui e z z
(29)
The amplitudes of each partial wave, B(− L) and B(−SV ) , are solved for by substituting into one of
the BCs:
σzz = 0 at z = 0, (30)
σxz = 0 at z = 0, (31)
and assuming B(− L)

= 1. This approach is taken in lieu of satisfying both BCs (by setting the
determinant of the coefficients matrix equal to zero), since k x is an approximate value, and the
determinant is never exactly zero, so most attempts at getting a reduced row echelon form of the
matrix result in the trivial solution of the eigenvector. To avoid this dilemma, we set B(− L) = 1,
and normalized as necessary. σxz = 0 could be used in place of the other BC, but only one is necessary.
Solving for B(−SV ) , it is found that B(−SV ) = −1.154. Plotting Equation (29) with the calculated values
of B(− L) and B(−SV ) gives the same wave structure for a Rayleigh wave as that found in Viktorov [9].
5.2. Stoneley Waves

The same approach can be applied to generalized Rayleigh waves. Generalized Rayleigh waves
refer to guided waves that demonstrate Rayleigh wave-like characteristics, such as for example,
Stoneley waves, Scholte waves, and Love waves [11–13]. For this example, Stoneley waves will be the
focus, but a similar approach can be applied to the others as well.
Stoneley waves, which propagate along the interface between two half-spaces, will be analyzed
as an example. Stoneley waves typically have a non-dispersive phase velocity that is slightly less
than the denser material’s bulk shear wave velocity, which means that it is in Region 1 (Figure 1a).
A Rayleigh wave-like wave structure should exist in the denser half-space [2]. A tungsten half-space
rigidly connected to an aluminum half-space will be used for the example. Using Auld’s example
(i.e., on pages 104–105 in [2]), the six eigensolutions are calculated for a frequency of 3.188 × 106 rad/s
and a wavenumber of 1172.1 rad/m. The frequency and wavenumber are chosen to meet the
non-dispersive phase velocity solution to the Stoneley wave problem [2]. The eigensolutions at
this frequency and wavenumber are shown in Tables 3 and 4 for tungsten and aluminum, respectively.
145
Appl. Sci. 2018, 8, 966
Table 3. Eigenvalues and normalized eigenvectors from the Christoffel equation for tungsten at
ω = 3.188 × 106 rad/s and k x = 1172.1 rad/m.

(+L) 997.6i −0.762i 0 0.648
(−L) −997.6i 0.762i 0 0.648
(+SV) 374.5i −0.304i 0 0.953
(−SV) −374.5i 0.304i 0 0.953
(+SH) 374.5i 0 1 0
(−SH) −374.5i 0 −1 0
ω = 3.188 × 106 rad/s and k x = 1172.1 rad/m.

(+L) 1058.1i −0.742i 0 0.671
(−L) −1059.1i 0.742i 0 0.671
(+SV) 562.5i −0.433i 0 0.902
(−SV) −562.5i 0.433i 0 0.902
(+SH) 562.5i 0 1 0
(−SH) −562.5i 0 −1 0
For both tungsten and aluminum, the Stoneley wave is in Region 1 (Figure 1a). To construct the
Stoneley wave displacement solution, the following partial waves are needed: solutions (− L) and
(−SV ) from aluminum’s Christoffel equation, and solutions (+ L) and (+SV ) from the tungsten’s
Christoffel equation. Solutions (− L) and (−SV ) from the aluminum are considered to be exponentially
decaying solutions for +z values, while solutions (+ L) and (+SV ) from the tungsten are considered
to be exponentially decaying solutions for −z values. These are shown in Figure 4 with respect to
the interface.
Figure 4. Conceptual depiction of the partial waves from tungsten and aluminum, which are necessary
for the Stoneley wave problem, overlaid on the two half-spaces.
The superposition of partial waves for each material, when ui (x, z, t) = ui (z)ei(−kx x+ωt) , is shown in
Equations (32) and (33). For tungsten (using values from Table 3):

(+ L) (+ L) −ik(+ L) (+SV ) −ik(+ SV )
i (z) = B
uW ui e z z + B(+SV ) ui e z z
(32)
while for aluminum (using values from Table 4):

(− L) −ik(− L) (−SV ) −ik(− SV )
uiAl (z) = B(− L) ui e z z + B(−SV ) ui e z z
(33)
146
Appl. Sci. 2018, 8, 966
Since the two half-spaces are assumed to be welded at the interface, the three of the four available
interface conditions:
Al
σzz = σzz
W
atz = 0, (34)
Al
σxz = σxz
W
atz = 0, (35)
uzAl = uW
z atz = 0, (36)
u xAl = uW
x atz = 0, (37)
can be used to solve for the amplitudes B(− L) , B(+SV ) , and B(−SV ) (i.e., assuming B(+ L) = 1).
Using these interface conditions, it was found that B(− L) = −0.3465, B(+SV ) = −1.2181,
B(−SV ) = 0.3108. Equations (32) and (33) can now be plotted to get the Stoneley wave’s wave structure,
as shown in Figure 5.
&RRUGLQDWH$ORQJWKH=D[LVP
Figure 5. Wave structure of a Stoneley wave calculated using the partial-wave method at
an aluminum/tungsten interface. Blue lines are the coordinate axis. −z is in the tungsten half-space,
and +z is in the aluminum half-space.
6. Lamb Waves and Shear Horizontal Waves

Consider a 1-mm thick aluminum plate with traction-free BCs, and the same frequency as was
analyzed in the previous section. The wavenumbers are calculated based on the Lamb wave dispersion
curve solutions at that frequency.
6.1. A0 Lamb Wave Mode

The six eigensolutions at a frequency of 3.188 × 106 rad/s and a wavenumber of 1688.7 rad/m
are shown in Table 5. This frequency–wavenumber pair corresponds with the red-triangle symbol
plotted in the dispersion-curve plot shown in Figure 2. Similar to the Christoffel equation solutions
used for the Rayleigh -wave case, the solutions in Table 5 are typical of Region 1 (Figure 1a), in which
the A0 mode exists. Unlike the Rayleigh wave case, there is now a bottom surface, so the exponentially
increasing eigensolutions are now necessary, although they will be written as exponentially decreasing
from the bottom surface upward as e−ikz (z− H ) . Using the coordinate system shown in Figure 6,
assignment to the top or bottom of the plate means that the partial waves will be defined using e−ikz (z)
or e−ikz (z− H ) , respectively. For simplicity, the rule used in this paper to categorize the partial waves
between the top or bottom of the plate is shown in Table 6.
147
Appl. Sci. 2018, 8, 966
ω = 3.188 × 106 rad/s and k x = 1688.7 rad/m.

(+L) 1612.4i −0.723i 0 0.691
(−L) −1612.4i 0.723i 0 0.691
(+SV) 1339.5i −0.621i 0 0.784
(−SV) −1339.5i 0.621i 0 0.784
(+SH) 1339.5i 0 1 0
(−SH) −1339.5i 0 −1 0
Table 6. Method for assigning partial waves to the top or bottom of the plate using the coordinate
system shown in Figure 6 and assuming e−ikz z .
Top of Plate Bottom of Plate

Real-valued (bulk waves) (+L) (+SV) (−L) (−SV)
kz and kz kz and kz
Imaginary-valued (surface waves) (−L) (−SV) (+L) (+SV)
kz and kz kz and kz
Figure 6. Conceptual depiction of the partial waves, which are necessary for the A0 Lamb wave
problem, overlaid on a plate in vacuum.
The superposed displacement profile in the z-direction is constructed by considering the symmetry of
the problem:
(+ L) −ik(+ L) (− L) −ik(− L) (+SV ) −ik(+ SV )

ui (z) = B(+ L) ui e z (z− H )
+ B(− L) ui ez z
+ B(+SV ) ui ez (z− H )
+
(38)
(−SV ) −ik(− SV )
B(−SV ) ui e z z
,
where ui (x, z, t) = ui (z)ei(−kx x+ωt) . For partial waves existing on the bottom surface of the plate, it is
shown in Equation (38) that the z-position is offset by the thickness of the plate. Just as in the Rayleigh
wave problem, B(−L) , B(+SV ) , and B(−SV ) can be solved by using three of the four BC available (assuming
that B(+L) = 1):
σzz = 0atz = 0, (39)
σxz = 0atz = 0, (40)
σzz = 0atz = H, (41)
σxz = 0atz = H, (42)
It was found that B(−L) = 1, and B(+SV ) = B(−SV ) = −1.0284. As before, Equation (38) can now be
used to plot the wave-structure of the A0 mode at this frequency.
148
Appl. Sci. 2018, 8, 966
6.2. S0 Lamb Wave Mode

The six eigensolutions at a frequency of 3.188 × 106 rad/s and a wavenumber of 591.57 rad/m are
shown in Table 7, where the shear vertical solutions have real-valued kz and real-valued eigenvectors.
This frequency–wavenumber pair corresponds with the red-diamond symbol plotted in the
dispersion-curve plot shown in Figure 2. These solutions correspond with the bulk waves reflecting
within the plate waveguide (see Figure 7).
ω = 3.188 × 106 rad/s and k x = 591.57 rad/m.

(+L) 312.93i −0.884i 0 0.468
(−L) −312.93i 0.884i 0 0.468
(+SV) 841.13 −0.818 0 0.575
(−SV) −841.13 0.818 0 0.575
(+SH) 841.13 0 1 0
(−SH) −841.13 0 −1 0
Figure 7. Conceptual depiction of the partial waves, which are necessary for the S0 Lamb wave problem,
overlaid on a plate in vacuum. The angle for the propagation direction of the shear vertical partial waves
was calculated using tan−1 (k x /kz ).
The displacement solution is constructed in a manner analogous to the A0 mode, except now there
are bulk waves present:
(+ L) −ik(+ L) (− L) −ik(− L) (−SV ) −ik(− SV )

ui (z) = B(+ L) ui e z (z− H )
+ B(− L) ui e z z
+ B(−SV ) ui e z (z− H )
+
(43)
(+SV ) −ik(+ SV )
B(+SV ) u i e z z
.
B(+L) , B(−SV ) and B(+SV ) are determined by using three of the four BCs (Equations (39)–(42))
(assuming that B(−L) = 1), and are found to be B(+L) = −0.998, B(−SV ) = −0.2139 + 0.4780i
and B(+SV ) = 0.2139 − 0.4779i. As before, Equation (43) can now be used to plot the wave structure of the
S0 mode at this frequency. Occasionally, the plotted wave structure will be at a phase that is inconsistent
with the conventional wave
structure
figures. This can be corrected by multiplying all of the displacements
imag(u )
by e−iφ , where φ = tan−1 real(u x) . This multiplication is equivalent to letting the wave propagate over
x
some non-zero distance and/or time.
149
Appl. Sci. 2018, 8, 966
6.3. Shear Horizontal Guided Wave Mode

Since the shear horizontal (SH) guided wave modes are similar to each other, a single discussion
of how each behaves is possible. The dispersion relationship found in Rose [4] is:
2
ωh
( Mπ )2 = − (kh)2 (44)
cS
M is the guided wave mode number, and h = H/2. The phase velocity is equal to the shear
wave velocity for M = 0, and for large ω values, it approaches the shear wave velocity for all M > 0.
The shear wave velocity represents the boundary between regions 1 and 2 (Figure 1a,b). That is, kz = 0
and k x = cω . The SH partial wave exists as a bulk wave in the plate that is propagating parallel to the
S
traction-free boundaries. When M > 0 and ω is not large, then the SH guided wave modes increase
in phase velocity and are highly dispersive. With the increasing phase velocity, the incidence angles of
the SH bulk waves decrease as the solution of the Christoffel equation traces the shear slowness curve
(the slowness curve representation of this dispersive SH guided wave mode is shown in Auld [2]).
7. The Contribution of Various Partial Waves in Each Lamb Wave
7.1. Symmetry and Opposing Phases

Based on the results from the S0 and A0 Lamb wave examples, it would seem as if there is an inherent
symmetry to the problem, since the amplitudes of the partial wave pairs often have equal magnitudes or
are out of phase by 180◦ . This is due to the problem’s symmetric geometry. This leads to the following
simplification to the displacement solution used to calculate the wave structure:

(± L) −ik(± L) (∓ L) −ik(∓ L)
ui ( z ) = B ( L) q ui e z (z− H ) + ui e z z
(45)
(±SV) −ik(± SV) (∓SV) −ik(∓ SV)
+ B(SV ) q ui e z (z− H )
+ ui e z z
where q is either +1 or −1. Determining which partial waves should be assigned to the top or bottom of
the plate is done in the same way as in the previous examples. As before, it is still implied by the use of the
Christoffel equation that ui (x, z, t) = ui (z)ei(−kx x+ωt) . In practice though, a few test calculations reveal that
although the amplitudes of the partial-wave pairs are close to equivalent, they are not exactly equivalent.
At best, Equation (45) can be used as a general description of the characteristics of many of the Lamb waves,
but it is not a mathematically viable solution for our current application. That being said, the general
description is invaluable for analyzing the characteristics of Lamb waves.
7.2. Using Symmetry to Summarize the Contribution of Longitudinal and Shear Partial Waves
One of the advantages of the partial-wave method approach is that the contribution of each partial
wave can be determined for each Lamb wave mode. That is, if the amplitudes are normalized by
dividing by:
2 2 2

2 1/2
Bmag = | B(+ L) | + | B(− L) | +| B(+SV ) | + | B(−SV ) | (46)
and since |B(+L) | ≈ |B(−L) | due to the symmetry of the problem, a representation of the amount of
longitudinal partial wave contribution can be calculated using the average of B(+L) and B(−L) :
( L) | B(+ L) | + | B(− L) |
Bavg = (47)
2Bmag
150
Appl. Sci. 2018, 8, 966
( L)
This average is used to account for the approximate nature of the solutions. Bavg is representative of
the amount of the longitudinal partial waves that make up a given Lamb wave mode, and is overlaid on
the dispersion curves as an intensity plot in Figure 8a.
Similarly, since |B(+SV ) | ≈ |B(−SV ) | due to the symmetry of the problem, the contribution of the shear
vertical partial waves can also be shown by calculating:
(SV ) | B(+SV ) | + | B(−SV ) |

Bavg = , (48)
2Bmag
and plotting it in Figure 8b.
(a) (b)
( L) (SV )
Figure 8. Amplitude of the (a) longitudinal (i.e., Bavg ) and (b) shear vertical (i.e., Bavg ) partial waves after
normalizing and averaging the magnitude for various points on the Lamb wave dispersion curves for
a 1-mm thick aluminum plate.
( L)
In Figure 8, it can be seen that in Region 1 (Figure 1a), the apportionment of longitudinal (i.e., Bavg )
(SV )
and shear vertical (i.e., Bavg ) partial waves are similar. However, in regions 2 and 3 (Figure 1b,c) however,
this is no longer the case. In Region 2 (Figure 1b), the Lamb waves around 4000 m/s are almost entirely
made up of bulk shear vertical partial waves, and Lamb waves just below 6000 m/s are made up of mostly
√
surface longitudinal partial waves. Interestingly, Lamb modes have phase and group velocities of 2cs
◦
and are comprised of shear bulk waves propagating at ±45 [14]. Region 2 (Figure 1b) also has a clear
phase velocity dependence on the amount of each partial wave, but no frequency dependence. In Region 3
(Figure 1c), a certain degree of mode dependence appears, but it does not appear to differentiate between
symmetric and anti-symmetric modes.
7.3. Using Opposing Phases to Sort Symmetric and Anti-Symmetric Lamb Wave Modes
Another advantage of using the partial-wave method is that the phase difference between the partial
waves at the top and bottom of the plate can be different depending on the type of Lamb wave. That is,
in symmetric modes, the top partial wave will be about 180◦ out-of-phase with the bottom partial wave.
Anti-symmetric modes will typically have the top and bottom partial waves in-phase. This can be
demonstrated by calculating q from Equation (45):
( L)
BTop
q= ( L)
(49)
BBottom
151
Appl. Sci. 2018, 8, 966
which should be equivalent to:

(SV )
BTop
q= (SV )
(50)
BBottom
However, since the frequency–wavenumber pairs are approximations, q will vary from −1 to +1
for most Lamb waves. The values of q are overlaid on the dispersion curves in Figure 9. It is worth
emphasizing that the phase difference between the top and bottom partial waves applies to both the
longitudinal and shear vertical partial waves, as shown in Equation (45).
Figure 9. q value which is indicative of the phase difference between the top and bottom partial waves for
various points on the Lamb wave dispersion curves for a 1-mm thick aluminum plate. The color axis is
constrained to +1 and −1.
Figure 9 demonstrates the phase difference characteristics of symmetric and anti-symmetric Lamb
waves, but it also shows how those characteristics break down at high frequencies where dispersion
curves intersect. At these intersection points, a phase difference (either +1 or −1) will dominate the
surrounding points of both dispersion curves. The clearest examples of this phenomenon are circled in red
in Figure 9.
8. Leaky Characteristics of Lamb Waves

Lamb waves travel through a traction-free plate. Leaky Lamb waves are the waves that propagate
through the same waveguide when the traction-free boundaries are replaced with fluid half-spaces.
The theoretical literature on leaky Lamb waves focuses on the calculation of dispersion curves and
determining how these curves differ from the conventional Lamb wave dispersion curves [15,16]. The focus
of this section is not this, but rather the leaky characteristics of Lamb waves, assuming there is no change
in the dispersion curves. That is, the dispersion curves of a plate surrounded by a vacuum will be
used as an approximation for the dispersion curves of a plate surrounded by air. This is a universal
standard practice. The goal will be to use the partial-wave method to determine the characteristics of the
acoustic leakage into the fluid from the Lamb wave travelling in the plate waveguide.
Before continuing, it should be stated that for fluids, the Eulerian form of the balance of linear
momentum for acoustics is:
∂ ∂ ∂
ρ vi + v j vi =− P (51)
∂t ∂x j ∂xi
This complicates the velocity–pressure relationship of the acoustic fluid. To simplify this, it is
assumed that the displacement is time-harmonic, which makes the particle velocity:
vi = iωui . (52)
152
Appl. Sci. 2018, 8, 966
It is also assumed that the displacement is spatially harmonic and longitudinal in polarization
such that:
ui ( x, y, z, t) = Uo k̂ i e−ik j x j +iωt , (53)
where k̂ j is the unit vector denoting the direction of wave propagation. Using Equations (52) and (53)
in Equation (51) results in:
iρ|vi |(ω − |k i ||vi |) = i |k i | P. (54)
The Eulerian form can approximate the Lagrangian form with these assumptions if:
ω |k i |ωUo . (55)
Equation (55) can be rearranged to:

cp
ω
. (56)
Uo
If it is assumed that the phase velocity is around 103 m/s and that displacements are about 10−6 m,
then for frequencies much less than 109 rad/s, the Eulerian form should approximate the Lagrangian form.
As a result, the simpler pressure–velocity relationship:
ρc L iωUo = P, (57)
which was derived from the Lagrangian form of the balance of linear momentum for fluids assuming
a time-harmonic displacement, can be used for the following example.
As an acoustic wave in a fluid, the slowness curve can be plotted as a circle with a radius equal
to 1/cair
L . Plotted with respect to aluminum’s slowness curve, the order of magnitude difference in radii
makes all but the very dispersive part of the A0 mode, at very low frequencies, emit acoustic bulk waves
(see Figure 10). At these low frequencies, the A0 mode is highly dispersive, and has a phase velocity lower
than the acoustic wave speed in air.
Figure 10. Depiction of the aluminum slowness curves compared with the slowness curve for air.
Slowness curves with solid lines denote real-valued kz solutions, and dotted lines denote imaginary-valued
kz solutions (figure not to scale).
The comparison of slowness curves from various media, which are shown in Figure 10,
is similar to an example in Auld [2], but that example discussed the electromagnetic emission from
a piezoelectric waveguide. The approach relies on Snell’s law, dictating that the wavenumber (i.e., k x )
153
Appl. Sci. 2018, 8, 966
should be equal at the interface between media. However, if Snell’s law were to be used directly,
complex-valued angles (i.e., θ) would need to be used.
Based on this slowness curve representation, it can be concluded that the acoustic leakage will be
an obliquely travelling bulk wave. The angle of incidence of the acoustic leakage would be between 0◦ and
20◦ for most Lamb waves. If these findings are applied to air-coupled transducers, it would suggest that
there is an optimum angle of reception for the air-coupled transducers.
The partial waves for air can be represented as an obliquely traveling longitudinal wave:
uiair ( x, z, t) = A air k̂ i e−ik j x j +iωt , (58)
and the partial waves of a Lamb wave in an aluminum plate (see Equations (38) and (43)) can be used
with five of the six available interface conditions:
P air = −σzz
Al
atz = 0, (59)
uzair = uzAl atz = 0, (60)

Al
σxz = 0atz = 0, (61)
P air
= −σzz
Al
atz = H, (62)
uzair = uzAl atz = H, (63)

Al
σxz = 0at z = H. (64)
where an expression for P air can be found by using Equations (57) and (58) to find that:
P air ( x, z, t) = iB air ρc L ωe−ik j x j +iωt (65)
Although stress components are expressly used in the boundary conditions detailed in this paper,
it is understood that they describe traction continuity. Equations (59)–(64) represent the traction and
displacement continuity conditions indicative of a fluid–solid interface. We assume that B(+L) = 1,
B(−L) , B(+SV ) , B(−SV ) , B(+air) , and B(−air) can be calculated as before by using the continuity conditions
in Equations (59)−(64). B(+air) and B(−air) in this case would correspond with the amplitudes of the
longitudinal waves at the top and bottom of the plate. All of the amplitudes were normalized by dividing
each by:
2 2 2 2 2

2 1/2
air
Bmag = | B(+ L) | + | B(− L) | + | B(+SV ) | + | B(−SV ) | +| B(+air) | + | B(−air) | (66)
Equations (38) and (58) can then be used to plot the wave structure, and verify that the interface
conditions are satisfied.
Since it is now possible to calculate the amplitude of the acoustic leakage, every point of the Lamb
wave dispersion curve can be input to determine the ideal guided wave mode for maximum (or minimum)
acoustic leakage. Figure 11 shows sample calculations for a 1-mm aluminum plate. The color axis of
Figure 11 is the average amplitude of acoustic leakage from Lamb waves:
( air ) | B(+air) | + | B(−air) |

Bavg = air
(67)
2Bmag
154
Appl. Sci. 2018, 8, 966
ܿ௅
ܿௌ
Figure 11. Amplitude of the longitudinal partial waves in the air half-space after normalizing and averaging
the magnitude for various points on the Lamb wave dispersion curves for a 1-mm thick aluminum plate.
The Lamb wave dispersion curves were calculated assuming traction-free boundary conditions (BC).
Assuming there is a direct correlation between the strength of acoustic leakage and the amplitude
of the acoustic wave coupled to the Lamb wave, Figure 11 suggests several things:
• the A0 mode can have a lower amplitude of acoustic leakage than the S0 mode (when normalized by
the amplitude of the wave structure);
• the amount of acoustic leakage is not mode-dependent;
• comparisons to Figure 8 suggest a potential relationship between the dominant partial waves present
and the amplitude of the acoustic leakage;
• when c p = cL , the leakage of the symmetric modes shown in Figure 11 agrees with the previous
analysis by Rose [4](Section 8.7.1), which attributes the lack of leakage to the lack of out-of-plane
displacement at that surface.
Perhaps most interesting is that the maximum acoustic leakage is concentrated in Region 2
(Figure 1b) of the dispersion curves, while Region 1 (Figure 1a) has noticeably less acoustic leakage.
A possible explanation for why this would be is that the guided wave modes in Region 1 (Figure 1a)
are composed of surface partial waves, which are inherently coupled to the interface.
In Region 2 (b), bulk shear vertical partial waves propagate through the thickness of the plate.
Incident upon the aluminum–air interface, this can be modeled as an oblique incidence problem to
determine the reflection and transmission characteristics, as shown in Figure 12 where θi is the incident
angle of a shear vertical bulk wave.
Figure 12. Depiction of the oblique plane wave problem being considered.
155
Appl. Sci. 2018, 8, 966
Displacement terms for each bulk wave are constructed according to particle polarization and
wave propagation direction. The same BCs that are detailed in Equations (59)−(64) are applied to
this problem. Along with Snell’s law, these BCs are used to calculate the displacement amplitude ratios
as a function of incidence angle, θi , as shown in Figure 13:
• R L is the particle displacement amplitude ratio between the reflected longitudinal wave in the
aluminum and the incident shear wave in the aluminum.
• RS is the particle displacement amplitude ratio between the reflected shear wave in the aluminum
and the incident shear wave in the aluminum.
• TL is the particle displacement amplitude ratio between the transmitted longitudinal wave in the
air and the incident shear wave in the aluminum.
( L) ( L) ( L)
It is noted that in place of sin θ, the ratio of wavenumbers were used, e.g., sin θ Al = k x /|k i |.
Figure 13. (a) The absolute value, (b) real part, and (c) imaginary part of the amplitude ratios for
an incident shear vertical bulk wave at an aluminum–air interface.
By setting up Snell’s law to be an equivalence between a bulk wave and an arbitrary wave in the
same plate, the relation is:
|ki |
cS−1 sin θi = sin θ. (68)
ω
Considering that θ is defined with respect to the z-axis, Snell’s law becomes:
c−
S sin θi = k x /ω = c p .
1
(69)
The black vertical line in Figure 13 corresponds with an incidence angle of about 45.45◦ ,
which translates to a phase velocity of 4350 m/s by Snell’s law. For reference, a phase velocity
156
Appl. Sci. 2018, 8, 966
at the shear wave speed, 3100 m/s, corresponds with an incidence angle of 90◦ , and a phase velocity at
the longitudinal wave speed, 6350 m/s, corresponds with an incidence angle of 29◦ . The angles in
the vicinity of 45.45◦ are associated with very little mode conversion to longitudinal modes, but most
importantly to a large transmission of acoustic longitudinal waves to the air half-space. The behavior
of the waves within the plate is consistent with what is Region 2 of Figure 8b.
Region 3 (c) would correspond to smaller angles of incidence as the phase velocity increases.
For angles of incidence that are less than the first critical angle (i.e., k x /ω < 1/c L ), the mode conversion
to longitudinal waves is significant with progressively less transmission of acoustic leakage as the
angle of incidence decreases. At the larger phase velocities in Region 3 (Figure 1c) (i.e., c p > c L ),
the longitudinal bulk waves traveling through the plate would then begin to contribute significantly to
any acoustic leakage.
In many studies of leaky Lamb waves, the fluid medium in which the wave leaks is water; thus,
it is worth mentioning that the dispersion curves do not change significantly when water coupling is
included [15,16]. As a result, many of the conclusions made in this example calculation pertaining to
the form of the acoustic leakage can be found to agree, to some degree, with the findings in Hayashi and
Inoue’s paper [15]. Specifically, the findings that the oblique propagation of the longitudinal acoustic
wave in the water and the evanescent propagation of it at a phase velocity less than the longitudinal
wave speed in water agree. We reiterate that Hayashi and Inoue [15] and Gravenkamp et al. [16] focus
on the study of the dispersion curves of leaky Lamb waves, which is different from what was done in
this example calculation.
9. Conclusions
This paper exploits the Christoffel equation, slowness curves, and the partial-wave method
to establish a foundation on which guided waves can be analyzed, compared, and explained.
The approach was used to calculate the wave structures of many types of guided waves, once the
frequency–wavenumber pairs were known. It is, of course, possible to calculate the wave
structures using the well-known methods for dispersion curve calculation, but those methods can
often be calculation-intensive. There are analytical expressions, but those leave little room for
conceptual interpretation. Thus, the approach detailed in this paper is seen as a relatively easy, general,
and informative method for calculating wave structures of specific modes of interest. Indicative of that
are the Lamb wave characteristics observed in the paper, which relate to the phase difference between
the top and bottom partial waves, and the amount by which each partial wave contributes to each
Lamb wave. It was shown, among other things, that in Region 2 (Figure 1b), there is a phase velocity
range in which the Lamb waves consist of partial shear vertical bulk waves traveling at ±45◦ . These are
known as Lame waves, and they exhibit the most leakage into the fluid half-spaces surrounding
the plate. Likewise, there is a similar band where all of the modes are dominated by longitudinal
surface waves. It was also shown that the symmetric and anti-symmetric modes can be characterized
by the phase difference between the partial waves associated with the top surface and those associated
with the bottom surface.
The example calculation of the leaky characteristics of Lamb waves is applicable to air-coupled
transducers. The optimum angle calculation is merely Snell’s law, but the calculation of the amplitude
of acoustic leakage provides us with a potentially simple methodology to compare the amount of
acoustic leakage from various Lamb waves. Also, the potential relationship between the acoustic
leakage from Lamb waves and the well-known oblique incidence problems could help to explain
some of the acoustic leakage phenomenon using familiar and relatively simple theories. Beyond Lamb
waves, a similar approach can be used for Rayleigh waves.
For other guided waves such as the Lamb-like waves in a plate on a half-space, the method
detailed in this paper should still apply. That is, there will be six partial waves (i.e., four in the plate
and two in the half-space) and six BCs (i.e., two traction-free at the top of the plate, and two traction
continuity and two displacement continuity at the interface).
157
Appl. Sci. 2018, 8, 966
Lastly, the partial-wave method is not at all restricted to isotropic materials; it can also be used
with anisotropic materials. If a similar approach is attempted for anisotropic materials, the immediately
discernable differences would be that there are four regions instead of three in the dispersion curves,
and three slowness curves instead of two. Some symmetry in the eigensolutions of the Christoffel
equation may also disappear as a result of the anisotropy. That being said, the general approach is still
expected to hold true, although further analysis is needed to confirm this.
Author Contributions: C.H. conducted the modeling and drafted the manuscript. C.L. advised C.H. and helped
edit the manuscript.
References
1. Solie, L.; Auld, B. Elastic waves in free anisotropic plates. J. Acoust. Soc. Am. 1973, 54, 50–65. [CrossRef]
2. Auld, B. Acoustic Fields and Waves in Solids; Robert, E., Ed.; Krieger Publishing Company: Malabar, FL, USA, 1990.
3. Rozzi, T.; Mongiardo, M. Open Electromagnetic Waveguides; The Institution Engineering and Technology:
London, UK, 1997.
4. Rose, J.L. Ultrasonic Waves in Solid Media; Cambridge University Press: New York, NY, USA, 1999.
5. Lowe, M. Matrix Techniques for Modeling Ultrasonic Waves in Multilayered Media. IEEE Trans. Ultrason.
Ferroelectr. Freq. Control. 1995, 42, 525–542. [CrossRef]
6. Crandall, S.H. On the Use of Slowness Diagrams to Represent Wave Reflections. J. Acoust. Soc. Am. 1970, 47,
1338–1342. [CrossRef]
7. Henneke, E.G. Reflection-Refraction of a Stress Wave at a Plane Boundary between Anisotropic Media.
J. Acoust. Soc. Am. 1972, 51, 210–217. [CrossRef]
8. Achenbach, J. Wave Propagation in Elastic Solids; North Holland Elsevier: New York, NY, USA, 1973.
9. Viktorov, I. Rayleigh and Lamb Waves: Physical Theory and Applications; Springer Science+Business Media:
New York, NY, USA, 1967.
10. Blackstock, D. Fundamentals of Physical Acoustics; John Wiley and Sons, Inc.: New York, NY, USA, 2000.
11. Stoneley, R. Elastic Waves at the Surface of Separation of Two Solids. R. Soc. Proc. 1924, 106, 416–428.
[CrossRef]
12. Scholte, J. The Range of Existence of Rayleigh and Stoneley Waves. Geophys. J. Int. 1947, 5, 120–126.
[CrossRef]
13. Scholte, J. On the Stoneley-wave equation. I. Proc. Acadamy Sci. Amst. (KNAW) 1942, 45, 20–25.
14. Matsuda, N.; Biwa, S. Phase and group velocity matching for cumulative harmonic generation in Lamb waves.
J. Appl. Phys. 2011, 109, 094903. [CrossRef]
15. Hayashi, T.; Inoue, D. Calculation of leaky Lamb waves with a semi-analytical finite element method.
16. Gravenkamp, H.; Birk, C.; Song, C. Numerical modeling of elastic waveguides coupled to infinite fluid
media using exact boundary conditions. Comput. Struct. 2014, 141, 36–45. [CrossRef]
158
applied
sciences
Article
Lamb Wave Local Wavenumber Approach for
Characterizing Flat Bottom Defects in an
Isotropic Thin Plate
Guopeng Fan 1 , Haiyan Zhang 1, *, Hui Zhang 1 , Wenfa Zhu 1,2 and Xiaodong Chai 2
1 School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China;
phdfanry@shu.edu.cn (G.F.); zh6154@126.com (H.Z.); zhuwenfa1986@163.com (W.Z.)
2 School of Urban Railway Transportation, Shanghai University of Engineering Science,
Shanghai 201620, China; xdchai@sues.edu.cn
* Correspondence: hyzh@shu.edu.cn; Tel.: +86-021-6613-7262
Received: 28 June 2018; Accepted: 24 July 2018; Published: 10 September 2018
Abstract: This paper aims to use the Lamb wave local wavenumber approach to characterize
flat bottom defects (including circular flat bottom holes and a rectangular groove) in an isotropic
thin plate. An air-coupled transducer (ACT) with a special incidence angle is used to actuate the
fundamental anti-symmetric mode (A0). A laser Doppler vibrometer (LDV) is employed to measure
the out-of-plane velocity over a target area. These signals are processed by the wavenumber domain
filtering technique in order to remove any modes other than the A0 mode. The filtered signals are
transformed back into the time-space domain. The space-frequency-wavenumber spectrum is then
obtained by using three-dimensional fast Fourier transform (3D FFT) and a short space transform,
which can retain the spatial information and reduce the magnitude of side lobes in the wavenumber
domain. The average wavenumber is calculated, as a real signal usually contains a certain bandwidth
instead of the singular frequency component. Both simulation results and experimental results
demonstrate that the average wavenumber can be used not only to identify shape, location, and size
of the damage, but also quantify the depth of the damage. In addition, the direction of an inclined
rectangular groove is obtained by calculating the image moments under grayscale. This hybrid and
non-contact system based on the local wavenumber approach can be provided with a high resolution.
Keywords: Lamb wave; local wavenumber; air-coupled transducer; wavenumber domain filtering;
hybrid and non-contact system
1. Introduction
Lamb waves have shown great potential for structural health monitoring (SHM) in plate-like
structures [1]. Their attractive features include sensitivity to a variety of damage types and the
capability of traveling relatively long distances [2]. Damage imaging methods based on Lamb waves
have been widely studied by many researchers, such as delay-and-sum (DAS) imaging [3,4], time
reversal focusing imaging [5,6], diffraction tomography imaging [7], and ultrasonic phased array [8].
Lamb waves are dispersive and multimodal [9]. Moreover, the propagating Lamb waves may include
incident, reflected, and converted waves when they encounter a sudden thickness variation [10,11], such
as circular flat bottom holes and rectangular grooves. Obviously, various wave modes make the
interpretation of Lamb waves very difficult [12].
In previous literature, the mode identification and separation of Lamb waves have been widely
investigated in the time domain, time-frequency domain, and frequency-wavenumber domain [13,14].
Short-time Fourier transform (STFT) with a sliding window can make the variant distribution of the
energy spectrum act as a function of time, which is capable of identifying Lamb wave mode [15].

Appl. Sci. 2018, 8, 1600
Frequency-wavenumber analysis based on two-dimensional fast Fourier transform (2D FFT) is a useful
approach to distinguish multiple modes. Moreover, the individual Lamb mode can be separated by a
2D band-pass filter [16]. It should be noted that the spatial information is also lost [17].
Compared with the above methods, the local wavenumber approach combined with a spatial
window is viewed as an effective way to obtain the space-frequency-wavenumber spectrum.
Furthermore, the spatial information can be reserved and the local features of the material can be
clearly characterized, including, for instance, the thickness distribution of the material.
Many researchers have tried to utilize the above techniques for SHM. Yu et al. adopted a short
space 2D Fourier transform to obtain the frequency-wavenumber spectrum at various spatial locations.
Consequently, a through-thickness crack of an aluminum plate was discriminated by analyzing the
space-frequency-wavenumber spectrum [18], but the size of the crack was not quantified. In another
study, Yu et al. proved that spatial wavenumber imaging method was able to characterize the location
and length of a crack. However, multiple cracks and cracks with various orientations were not
researched in their work [19]. Rogge et al. utilized the local wavenumber analysis to characterize
the impact damage in composite laminates [20]. The location of a preset delamination was clearly
characterized, while its depth was not quantified.
Rao et al. reconstructed the thickness distribution of a liquid loaded plate by using an ultrasonic
guided wave tomography method based on full waveform inversion (FWI) [21,22]. It was noted that
the remaining wall thickness was reconstructed with high accuracy and resolution. Chronopoulos
et al. adopted an inverse wave and finite element approach to recover the thickness and density, as
well as all independent mechanical characteristics of layered composite structures [23]. However,
one drawback of this method is that it is not efficient, as the inversion process is complicated and
time-consuming. Therefore, a rapid method for SHM is desired.
Air-coupled ultrasonic inspection is a promising non-contact method for the rapid, non-contact
inspection of materials [24,25]. The main advantages of this technique are the absence of any contact
and the ability to generate a relatively pure Lamb wave mode when an appropriate incidence angle is
adopted. The laser Doppler vibrometer (LDV) based on the Doppler effect has a high spatial resolution
in obtaining the propagating Lamb waves [26]. Harb and Yuan accomplished the target of actuating
the pure A0 mode (anti-symmetric mode) in an isotropic aluminum plate by employing an air-coupled
transducer (ACT) and an LDV [27].
A hybrid non-contact system composed of an ACT and an LDV is presented in this paper. The local
wavenumber approach is adopted to acquire the average wavenumber distribution of an isotropic
thin plate, which contains circular flat bottom holes or a rectangular groove. The main goal is not only
to extract defect features in location, size, and shape, but also quantify the depth of the damage by
combining the wavenumber with the phase velocity dispersion curve. The direction of a rectangular
groove can be attained by calculating the image moments under grayscale. Several factors which can
affect test quality are analyzed in the discussion section. Moreover, the solutions are presented when
the proposed approach is applied for an unknown damage.
This paper is organized with seven sections, including this introduction. Section 2 illustrates
the theory of the local wavenumber approach. Section 3 presents the results of tests performed on
numerically simulated examples. Experimental results are explicated in Section 4. Section 5 concludes
the discussion. Section 6 provides recommendation for future research. Section 7 summarizes the
main work.
2. Theory for the Local Wavenumber Approach
2.1. Sound Field Transmission and Reception

The A0 mode is the appropriate Lamb wave mode for characterizing circular flat bottom holes
and a rectangular groove in an isotropic thin plate, as the phase velocity of the A0 mode is sensitive to
the thickness-frequency product. First of all, a special incidence angle can be easily calculated based
160
Appl. Sci. 2018, 8, 1600
on Snell’s law. Then, the fundamental anti-symmetric mode (A0) can be actuated by an ACT with this
special incidence angle. A scanning grid is predefined before using a non-contact LDV to measure the
velocities of the desired mode. Obviously, the velocity U ( x, y, t) is a function of both time and space.
Figure 1 shows the theoretical dispersion curves obtained by a commercial software package called
Disperse (version 2.0.20a, Imperial College London, London, UK) [28]. In this paper, the incidence
angle is set at 12.75◦ , which is suitable for an ACT to actuate the A0 mode in an aluminum plate with
the frequency-thickness product of 0.3 MHz·mm. Figure 2 shows a schematic diagram, which is used
for describing the process of sound field transmission and reception.
Figure 1. Dispersion curves of an aluminum plate. (a) Phase velocity; (b) incidence angle;
(c) wavenumber.
Figure 2. Schematic diagram of Lamb wave transmission and reception.
2.2. Identification and Separation of Lamb Wave Modes in the Wavenumber Domain
Only the fundamental anti-symmetric mode (A0) can be actuated by ACT with the special
incidence angle; however, mode conversion still exists due to the interaction between the Lamb wave
and the flat bottom defects. The received Lamb wave signals usually comprise the incident, reflected,
and converted wave modes. By using three-dimensional fast Fourier transform (3D FFT) [29], the full
wavefield data can be transformed into the wavenumber domain, where various wave modes can
be distinguished.
U (k x , k y , f ) = F3D (U ( x, y, t)) (1)
161
Appl. Sci. 2018, 8, 1600
The entire wavefield signals are filtered by the wavenumber domain filter for the purpose of
removing any modes other than the A0 mode.
(k x , k y , f ) = Wk (k x , k y , f )U (k x , k y , f )
U (2)
⎧
⎪
⎨ 0 |k| ≤ k1
Wk (k x , k y , f ) = 1 k1 < |k| < k2 (3)
⎪
⎩ 0 |k| ≥ k2

kx
k= (4)
ky
These filtered signals in the wavenumber domain should be transformed back into the time
domain by using three-dimensional inverse fast Fourier transform (3D IFFT). U ( x, y, t) is the
filtered wavefield.
U (k x , k y , f ))
( x, y, t) = F3D −1 (U (5)
2.3. Acquisition of the Average Wavenumber

A spatial window is used for reducing the magnitude of the side lobes. It should be noted that
the spatial window size is at least twice the wavelength of the A0 mode in this paper. The centered
coordinate of this spatial window is located at ( xm , yn ). Under this circumstance, the process of
obtaining the average wavenumber is clearly illustrated as below.
First, a short space transform is conducted in order to retain the spatial information and reduce
the magnitude of side lobes in the wavenumber domain. Specifically, the wavefield data is multiplied
by a spatial window to produce the windowed dataset U mn ( x, y, t). The spatial window is non-zero
for only a short period in space. The Hanning window Wxy is usually adopted [30].
mn ( x, y, t) = U
U ( x, y, t)Wxy (6)

0.5 1 + cos(2π Drx ) if r ≤ Dx /2
Wxy = (7)
0 otherwise
where r and Dx can be expressed as:

r= ( x − x m )2 + ( y − y n )2 (8)
Dx ≥ 2λ (9)
Here, λ denotes the wavelength of the Lamb wave. Dx is the window length. These signals
processed with the spatial window are transformed into the spatial-frequency-wavenumber domain
by using 3D FFT.
U mn ( x, y, t)]
mn (k x , k y , f ) = F3D [U (10)
The weight of k must be considered before calculating the local wavenumber k mn , as presented
in the following equation: !
2

∑ U ( k
mn x y, k , f ) k
k mn ( f ) = k 2 (11)

∑ U mn ( k x , k y , f )
k
where k is the Euclidean norm of the 2D wavenumber range [31]. The average wavenumber k mn over
the selected frequency band is calculated because a real signal usually contains a certain bandwidth
instead of the singular frequency component. The average wavenumber of each point can be obtained
162
Appl. Sci. 2018, 8, 1600
by changing the centered coordinate of this spatial window and repeating the above steps. The flow
chart of the local wavenumber approach is shown in Figure 3.
1 N
N i∑
k mm = k mn ( f i ) (12)
=1
Figure 3. Flow chart of the local wavenumber approach.
2.4. Depth Reconstruction

The average wavenumber k mn at each point can be obtained by the local wavenumber approach.
Hence, the phase velocity C p ( xm , yn ) can be calculated as below:
2π f
C p ( xm , yn ) = (13)
k mn
The relationship between the phase velocity and the plate thickness can be gained from the phase
velocity dispersion curve shown in Figure 1a. Consequently, the plate thickness d( xm , yn ) of the target
area can be reconstructed according to this relationship. Then, the depth distribution can be written as:
h( xm , yn ) = D − d( xm , yn ) (14)
where D is the thickness of the pristine plate. In order to evaluate the depth, the maximum depth error
Edepth is defined as below:
|hmax − h0 |
Edepth = (15)
h0
Here, hmax denotes the maximum value of the reconstructed depth and h0 is the actual value.
3. Finite Element Simulation
3.1. Finite Element Model

A commercial software package called PZFlex was used to create the finite element model, which
is shown in Figure 4. PZFlex is the time domain finite element method (FEM). The complex acoustic
wave propagation in a three-dimensional model can be simulated by using this method [32]. The upper
and bottom surfaces of the model are free, while the remaining four sides are absorbing. The size of the
6061-T6 aluminum plate is 400 mm × 400 mm × 1.5 mm, and the other material properties are listed
in Table 1. The incident point of the ACT is (100, 200), and its incidence angle is set at 12.75◦ , which is
suitable to actuate the A0 mode on aluminum with the frequency-thickness product of 0.3 MHz·mm.
The damage located at the middle of the plate is a circular flat bottom hole with a diameter of 10 mm
and a depth of 1 mm. This damage is surrounded by a rectangular scanning area (60 mm × 60 mm),
163
Appl. Sci. 2018, 8, 1600
which is divided into 121 × 121 points with a spatial interval of 0.5 mm. The spatial interval should
not be more than half of the wavelength according to the sampling theorem. In our test, the phase
velocity of the A0 mode is about 1551 m/s, and the central frequency of the received signal is 200 kHz.
Therefore, the wavelength of the A0 mode is 7.775 mm (λ = 1511/200 = 7.775 mm). The sampling
interval is only 0.5 mm (λ/15). Apparently, the sampling interval satisfies the sampling theorem and
can provide a good spatial resolution. The out-of-plane velocities over the scanning area are recorded
by the LDV.
In this paper, a five-cycle sinusoidal tone burst modulated by a Hanning window with a central
frequency of 200 kHz is applied as the excitation signal, which is shown in Figure 5. Eleven specimens
listed in Table 2 are used to evaluate the performance of the local wavenumber approach from
different aspects.
Case 1 and Case 2 are used to evaluate the performance of the proposed method when the circular
flat bottom hole changes in depth.
Case 1 and Case 3 are used to evaluate the performance of the proposed method when the circular
flat bottom hole changes in radius.
Case 4 to Case 7 are used to study the impacts caused by the excitation source position and the
distance among adjacent defects. The excitation source in Case 4 and Case 5 is located at (100, 200),
while it is located at (200, 100) in Case 6 and Case 7. The gap between the two flat bottom holes is
20 mm in Case 4 and Case 6, while it is 10 mm in Case 5 and Case 7.
Case 8 and Case 9 are used to investigate whether the impact which the big hole exerts on the
small one is connected with their positions relative to the excitation source when they are in the
same line.
Case 10 and Case 11 are used to identify the direction of the rectangular groove.
Table 1. Properties of the 6061-T6 aluminum plate.
Property Value
Density 2700 (kg/m3 )
Young’s modulus 68.9 (GPa)
Poissons’ratio 0.33
Figure 4. Finite element model.
164
Appl. Sci. 2018, 8, 1600
Figure 5. The excitation signal in the (a) time-domain; (b) frequency-domain.
Table 2. Specimen parameters.
Excitation
Central Position Radius Depth Inclined Width Length
Specimen Damage Type Source
of Damage (mm) (mm) (mm) Degree (◦ ) (mm) (mm)
Position (mm)
Case 1 One flat bottom hole (100, 200) (200, 200) 5 1 / / /
Case 2 One flat bottom hole (100, 200) (200, 200) 5 0.5 / / /
Case 3 One flat bottom hole (100, 200) (200, 200) 3.5 1 / / /
No. 1: (200, 215) 5 1 / / /
Case 4 Two flat bottom holes (100, 200)
No. 2: (200, 185) 5 1 / / /
No. 1: (200, 210) 5 1 / / /
No. 2: (200, 190) 5 1 / / /
No. 1: (200, 215) 5 1 / / /
No. 2: (200, 185) 5 1 / / /
No. 1: (200, 210) 5 1 / / /
No. 2: (200, 190) 5 1 / / /
No. 1: (200, 210) 5 1 / / /
No. 2: (200, 190) 3.5 1 / / /
No. 1: (200, 210) 3.5 1 / / /
No. 2: (200, 190) 5 1 / / /
Case 10 One groove (100, 200) (200, 200) / 1 90 4 24
Case 11 One groove (100, 200) (200, 200) / 1 135 4 24
3.2. FEM Results
3.2.1. Mode Separation

The interactions between Lamb waves and the damage have been analyzed by many researchers.
Bhuiyan et al. adopted the wave damage interaction coefficient (WDIC) to investigate the scattering of
Lamb waves and demonstrated that the damage in the plate acted as a non-axisymmetric secondary
source [33]. In another report, Bhuiyan et al. described the 3D interaction (frequency, incident direction,
and azimuth direction) of Lamb waves with the damage [34]. Testoni et al. adopted a wavenumber
filter to filter out the main wavefield generated by the acoustic source. Thus, the wave scattered by
the delamination can be clearly observed in the filtered signal [35]. In this paper, Case 1 is taken as
an example to describe the interaction between Lamb waves and the damage; moreover, the mode
separation is introduced in detail.
The snapshots of the Lamb waves are shown in Figure 6. It can be observed that only the A0 mode
can be actuated by an ACT with a special incidence angle. However, the S0 mode will be generated
when the A0 mode encounters a flat bottom defect. Following this, the incident A0, reflected A0,
transmitted A0, and S0 mode will coexist for a long time. It is worth noting that the amplitude of the
S0 mode is very low.
165
Appl. Sci. 2018, 8, 1600
Figure 6. Snapshots of Lamb waves. (a) 74.18 us; (b) 91.31 us; (c) 105.89 us; (d) 111.60 us; (e) 114.14 us;
(f) 120.48 us. The white circle denotes the incident point of the air-coupled transducer (ACT). The black
circle represents the actual location of a defect.
The Lamb wave propagation was analyzed in detail. The next section will be focused on
mode separation. The frequency-thickness product is 0.3 MHz·mm at the undamaged area. Under
this situation, the theoretical wavenumbers of the A0 mode and S0 mode are 0.8102 rad/mm
and 0.2312 rad/mm, respectively. The frequency-thickness product is only 0.1 MHz·mm at the
damage. Hence, the theoretical wavenumbers of the A0 mode and S0 mode are 1.3082 rad/mm and
0.2311 rad/mm, respectively. Apparently, the wavenumber of the A0 mode changes more obviously
than that of the S0 mode when the thickness changes. Figure 7 shows the mode separation by using
3D FFT and the wavenumber domain filtering technique. The wavenumber spectrum is obtained at a
frequency of 200 kHz. The images are given in dB scale so that the wavenumber spectrum can be clearly
visualized. The wavenumber spectrum for the pristine plate is shown in Figure 7a. The wavenumber
spectrum for the damaged plate is shown in Figure 7b. The high-pass filter in wavenumber domain
is shown in Figure 7c. The filtered wavenumber spectrum is shown in Figure 7d. By comparing the
wavenumber spectrum in Figure 7a,b, we can confirm the fact that the A0 mode has interacted with
the defect. In addition, the S0 mode generated by mode conversion is very weak, as the amplitude of
the S0 mode is much lower than that of the A0 mode in the time domain.
Generally, a band-pass filter should be applied to reserve the A0 mode according to the theory of
the local wavenumber approach. However, the wavenumber distribution of the A0 mode is wide in
our test, while the wavenumber distribution of the S0 mode is very narrow. Moreover, the intensity of
the S0 mode is weak. Therefore, it is easy to design a high-pass filter rather than a band-pass filter.
The high-pass filter should ensure that the wavenumbers of the A0 mode are included in the pass band.
166
Appl. Sci. 2018, 8, 1600
Figure 7. Mode separation based on the spatial-wavenumber filtering technique. (a) Wavenumber
spectrum for the pristine plate; (b) wavenumber spectrum for the damaged plate; (c) high-pass filter in
the wavenumber domain; (d) filtered wavenumber spectrum. White circles are theoretical wavenumber
curves for the S0 and A0 modes.
3.2.2. FEM Results of a Circular Flat Bottom Hole

In this paper, the wavelength of the A0 mode is 7.775 mm. Therefore, the window radius
(Dx /2) should be larger than 7.775 mm. It is assumed that the window radius is 8 mm. The average
wavenumber can be obtained using Equation (12). Figure 8 shows the acquisition of the average
wavenumber for Case 1. The frequency interval is 15 kHz. It can be seen that the average wavenumber
contributes to improve the test quality at the scanning boundary by comparing Figure 8d,h. Moreover,
lower image noise is seen in the image when the average wavenumber is adopted.
167
Appl. Sci. 2018, 8, 1600
Figure 8. The acquisition of average wavenumber for Case 1. (a) Wavenumber distribution obtained
at f = 155 kHz; (b) wavenumber distribution obtained at f = 170 kHz; (c) wavenumber distribution
obtained at f = 185 kHz; (d) wavenumber distribution obtained at f = 200 kHz; (e) wavenumber
distribution obtained at f = 215 kHz; (f) wavenumber distribution obtained at f = 230 kHz;
(g) wavenumber distribution obtained at f = 245 kHz; (h) average wavenumber distribution.
In fact, the best test result is not always obtained under the condition in which the window radius
is equal to the wavelength. It is meaningful to analyze the test results under various window radii in
order to find the optimum size. The test results shown in Figure 9 indicate that the damage region
characterized by the proposed approach will expand as the window radius increases. Meanwhile, the
maximum wavenumber attained under the various radii will decrease as the window radius increases,
which is shown in Figure 10. It is only the window radius of 8 mm that can make the identified damage
168
Appl. Sci. 2018, 8, 1600
region match well with its actual size. Therefore, 8 mm is the optimum size. All of the test results were
obtained under this condition.
Figure 9. The test results obtained under various window radii. (a) 8 mm; (b) 9 mm; (c) 10 mm;
(d) 11 mm; (e) 12 mm; (f) 13 mm. The white circle represents the actual location of the damage.
The central frequency of the excitation signal is 200 kHz.
Figure 10. The maximum wavenumber obtained under various window radii. The central frequency
of the excitation signal is 200 kHz.
169
Appl. Sci. 2018, 8, 1600
Case 1 and Case 2 only contain a circular flat bottom hole with the same radius and different
depths. Case 1 and Case 3 only contain a circular flat bottom hole with different radii and the same
depth. The specimen parameters are listed in Table 2. The simulation results of Case 1 to Case 3 are
shown in Figure 11. The size and shape of the defect can be clearly visualized. The depth estimations
are shown in Table 3. The depth error of the deep hole is 1.13%, while it reaches 3.26% for the shallow
hole. Apparently, the depth estimation of the deep hole is more accurate than that of the shallow one.
The depth error of the big hole is 1.13%, while it reaches 8.17% for the small one. Hence, the depth
estimation of the big hole is more accurate than that of the small one.
Figure 11. Simulation results of Case 1 to Case 3. (a) Wavenumber distribution of Case 1; (b) depth
distribution of Case 1; (c) wavenumber distribution of Case 2; (d) depth distribution of Case 2;
(e) wavenumber distribution of Case 3; (f) depth distribution of Case 3. The white circle represents the
actual location of the damage. The actual defect depth of Case 1 and Case 3 is 1 mm, while the actual
defect depth of Case 2 is 0.5 mm.
Table 3. The depth evaluations of Case 1 to Case 3.
Specimen Actual Depth (mm) Reconstructed Depth (mm) Estimation Error of Depth
Case 1 1 0.9887 1.13%
Case 2 0.5 0.4837 3.26%
Case 3 1 0.9183 8.17%
170
Appl. Sci. 2018, 8, 1600
3.2.3. FEM Results of Two Circular Flat Bottom Holes with the Same Radius and Depth
Case 4 to Case 7 contain two circular flat bottom holes with the same radius and depth, but the
excitation source location and the gap between two defects are different. The specimen parameters are
listed in Table 2. The simulation results of Case 4 to Case 7 are shown in Figure 12. It is observed that
the shape and location of the defects are recognizable. The two holes exhibit a symmetrical distribution
when the excitation source is located at (100, 200). In this condition, the better result appears under a
big gap between the defects. The two holes and the excitation source are in the same line when the
excitation source is located at (200, 100). In this condition, it can be noticed that the hole far away from
the excitation source seems smaller than the near one. Moreover, the closer the defects are, the more
obvious the effect is. Table 4 shows that the depth errors of Case 4 to Case 7 are all within 2%.
Figure 12. Simulation results of Case 4 to Case 7. (a) Wavenumber distribution of Case 4; (b) depth
distribution of Case 4; (c) wavenumber distribution of Case 5; (d) depth distribution of Case 5;
(e) wavenumber distribution of Case 6; (f) depth distribution of Case 6; (g) wavenumber distribution of
Case 7; (h) depth distribution of Case 7. The white circle represents the actual location of the damage.
The actual depth of the defect is 1 mm.
171
Appl. Sci. 2018, 8, 1600
Table 4. The depth evaluations of Case 4 to Case 7.
No. 1 1 0.9917 0.83%
Case 4
No. 2 1 0.9917 0.83%
No. 1 1 0.9887 1.13%
Case 5
No. 2 1 0.9887 1.13%
No. 1 1 0.9946 0.54%
Case 6
No. 2 1 0.9976 0.24%
No. 1 1 0.9887 1.13%
Case 7
No. 2 1 0.9976 0.24%
3.2.4. FEM Results of Two Circular Flat Bottom Holes with Different Radius and the Same Depth
Two circular flat bottom holes with different radii and the same depth are used in Case 8 and
Case 9. It is important to note that these defects and the excitation source are in the same line, which
means that the defects have the different positions relative to the excitation source. The specimen
parameters are listed in Table 2. Simulation results are shown in Figure 13. It can be seen that the shape
and location of two holes can be identified, which demonstrates that the local wavenumber approach
has a high resolution even in distinguishing adjacent defects. Table 5 shows the depth evaluations of
Case 8 and Case 9. The depth error of the big hole is within 1%, while the depth error of the small one
is in the range of 4% to 9%. Thus, the depth estimation of the big hole is more accurate than that of
the small one. Besides, the relative position is also taken into consideration. The depth error of the
small hole is 8.17% when it is in front of the big one. Apparently, this level is equivalent to that of
Case 3. However, the depth error of the small hole is reduced to 4.76% when it is behind the big one.
This indicates that the impact which the big hole exerts on the small one is related to their positions
when they and the excitation source are in the same line.
Figure 13. Simulation results of Case 8 and Case 9. (a) Wavenumber distribution of Case 8; (b) depth
distribution of Case 8; (c) wavenumber distribution of Case 9; (d) depth distribution of Case 9. The white
circle represents the actual location of the damage. The actual depth of the defect is 1 mm.
172
Appl. Sci. 2018, 8, 1600
Table 5. The depth evaluations of Case 8 and Case 9.
No. 1 1 0.9976 0.24%
Case 8
No. 2 1 0.9183 8.17%
No. 1 1 0.9524 4.76%
Case 9
No. 2 1 0.9917 0.83%
3.2.5. FEM Results of an Inclined Rectangular Groove

Case 10 is a rectangular groove with an inclined angle of 90◦ . Case 11 is a rectangular groove with
an inclined angle of 135◦ . The specimen parameters are listed in Table 2. The simulation results of Case
10 and Case 11 are shown in Figure 14. The damage shape can be clearly visualized. It should be noted
that the length of this defect is a little shorter than the actual value, while its width is obviously larger
than the actual value. In addition, the orientation of the rectangular groove can be easily obtained and
the computational process is introduced as below.
Figure 14. Simulation results of Case 10 and Case 11. (a) Wavenumber distribution of Case 10; (b) depth
distribution of Case 10; (c) wavenumber distribution of Case 11; (d) depth distribution of Case 11.
The white rectangle represents the actual location of the damage. The actual depth of the defect is 1 mm.
In image processing, computer vision, and related fields, an image moment is a certain particular
weighted average (moment) of the image pixels’ intensities. Image moments and moment invariants
are useful to describe global features in image processing [36]. Image moments are calculated under
grayscale in order to distinguish the direction of the rectangular groove. G ( x, y) denotes the pixel
intensity of the gray value, where x, y are the row and column of the pixel matrix, respectively. The gray
image must be processed by a threshold as we focus attention on the area where the rectangular groove
is likely to exist rather than the whole area. I ( x, y) is the pixel intensity obtained by setting a threshold
at the level of −3 dB. The zero moment is shown in Equation (16). The first moment is shown in
Equation (17). The second moment is shown in Equation (18). These image moments can be used to
extract image features [37].
M00 = ∑ ∑ I ( x, y) (16)
x y
173
Appl. Sci. 2018, 8, 1600
⎧
⎪
⎨ M10 = ∑ ∑ x × I ( x, y)
x y
(17)
⎪
⎩ M01 = ∑ ∑ y × I ( x, y)
x y
⎧
⎪
⎪ M20 = ∑ ∑ x2 × I ( x, y)
⎪
⎪ x y
⎨
M02 = ∑ ∑ y2 × I ( x, y) (18)
⎪
⎪ x y
⎪
⎪
⎩ M11 = ∑ ∑ x × y × I ( x, y)
x y
M00 , M10 , and M01 are used to obtain the centroid coordinates of the feature image [38].
The calculation formulas are written as below.
⎧ M10
⎪
⎨ x=
M00 (19)
⎪
⎩ y = M01
M00
The spindle orientation of the feature image can be obtained by calculating an angle, which is
shown in the following equation.
/ ( μ − μ ))
arctan(2μ11
Θ= 20 02
(20)
2
, μ , and μ can be given by the following equations:
where μ11 20 02
⎧
⎪
⎪ = M20 − x2
μ20
⎪
⎪
⎪
⎨
M00
M02
μ02 = − y2 (21)
⎪
⎪ M00
⎪
⎪
⎪
⎩ μ = M11 − x × y
11 M00
Figure 15 shows the orientation of an inclined rectangular groove. The depth and angle
evaluations are shown in Table 6. It is easy to find that the angle errors of Case 10 and Case 11
only reach 0.00% and 0.12%, respectively, while the depth errors of Case 10 and Case 11 rise to 12.95%
and 13.60%, respectively. One reasonable explanation is clarified in the discussion section.
Figure 15. The direction identification of an inclined rectangular groove based on the numerical result.
(a) 90◦ ; (b) 135◦ . The red plus (+) indicates the centroid location. The blue dashed line is the spindle.
The green dash line is the boundary of the inclined rectangular groove. The background image is a
grayscale image which is processed by a threshold at the level of −3 dB.
174
Appl. Sci. 2018, 8, 1600
Table 6. The depth and angle evaluations of Case 10 and Case 11.
Actual Reconstructed Estimation Error Actual Incidence Reconstructed Estimation Error

Specimen
Depth (mm) Depth (mm) of Depth Angle (◦ ) Angle (◦ ) of Angle
Case 10 1 0.8705 12.95% 90 90.00 0.00%
Case 11 1 0.8640 13.60% 135 134.84 0.12%
4. Experimental Verification
4.1. Experimental Setup

The experimental system is composed of an ACT, an LDV, an ultrasonic power device, a signal
amplifier, a digital phosphor oscilloscope, and a data acquisition interface. The ACT is a piezoelectric
ceramic-based transducer with a 200 kHz center frequency, and a bandwidth of ±25% of the center
frequency. The incidence angle is set at 12.75◦ , which is suitable for the ACT to actuate the A0 mode.
The LDV is employed to measure the out-of-plane velocity on the surface of the plate. The sampling
frequency is set at 500 MHz. The signal should be executed at a suitable time when the waveform
displayed on the oscilloscope tends to be stable. Each measurement is repeated 10 times in order to
reduce the measurement error. The received signals need be amplified by a wideband amplifier for the
purpose of improving the signal-to-noise ratio. The gain of the signal amplifier is 60 dB. Moreover, it can
be adjusted according the central frequency of the excitation signal. The data acquisition subsystem
provides detailed information of the recording signals, such as the sample interval, trigger point,
vertical scale, horizontal scale, etc. The experimental system is shown in Figure 16. The experimental
conditions are in agreement with the simulation conditions.
Figure 16. Schematic of the experimental setup.

For the purpose of removing the noise, the received signals are filtered by a Chebyshev band-pass
filter [39] with a bandwidth of 140 kHz to 260 kHz. Figure 17 shows the experimental signal in the
time domain and frequency domain. Figures 18–22 are the experimental results.
175
Appl. Sci. 2018, 8, 1600
Figure 17. Band-pass filtering of an experimental signal. (a) Original signal; (b) spectrum of the original
signal; (c) filtered signal; (d) spectrum of the filtered signal.
Figure 18. Experimental results of Case 1 to Case 3. (a) Wavenumber distribution of Case 1; (b) depth
distribution of Case 1; (c) wavenumber distribution of Case 2; (b) depth distribution of Case 2;
(d) wavenumber distribution of Case 3; (e) depth distribution of Case 3. The white circle represents the
actual location of the damage. The actual defect depth of Case 1 and Case 3 is 1 mm, while the actual
defect depth of Case 2 is 0.5 mm.
176
Appl. Sci. 2018, 8, 1600
Figure 19. Experimental results of Case 5 and Case 7. (a) Wavenumber distribution of Case 5; (b) depth
Figure 20. Experimental results of Case 8 and Case 9. (a) Wavenumber distribution of Case 8; (b) depth
177
Appl. Sci. 2018, 8, 1600
Figure 21. Experimental results of Case 10 and Case 11. (a) Wavenumber distribution of Case 10;
(b) depth distribution of Case 10; (c) wavenumber distribution of Case 11; (d) depth distribution of
Case 11. The white rectangle represents the actual location of the damage. The actual depth of the
defect is 1 mm.
Figure 22. The direction identification of an inclined rectangular groove based on the experimental
result. (a) 90◦ ; (b) 135◦ . The red plus (+) indicates the centroid location. The blue dashed line is the
spindle. The green dash line is boundary of the inclined rectangular groove. The background image is
a grayscale image which is processed by a threshold at the level of −3 dB.
The experimental results have good agreement with the simulation results. Figure 18 shows that
the depth estimation of the deep hole is more accurate than that of the shallow one. In addition, the
depth estimation of the big hole is more accurate than that of the small one. Figure 19 reveals that the
circular flat bottom hole far away from the excitation source seems smaller than the near one when the
excitation source and two identical defects are in the same line. Figure 20 shows that the proposed
approach is capable of identifying the adjacent defects with different sizes. Meanwhile, it indicates
that the impact that the big hole exerts on the small one is connected to their positions relative to the
excitation source when they are in the same line. Figure 21 shows that the width of the rectangular
groove is larger than its actual value. Figure 22 shows the direction identification for an inclined
rectangular groove. Tables 7 and 8 indicate that the evaluation results still need to be improved due to
the existence of multi-reflection between the adjacent defects and the spatial window effect, which will
be analyzed in the discussion section.
178
Appl. Sci. 2018, 8, 1600
Table 7. The depth evaluations of Case 1 to Case 3, Case 5, Case 7, Case 8, and Case 9.
Case 1 1 0.9432 5.68%
Case 2 0.5 0.4490 10.20%
Case 3 1 0.8673 13.27%
No. 1 1 0.9308 6.92%
Case 5
No. 2 1 0.9401 5.99%
No. 1 1 0.9152 8.48%
Case 7
No. 2 1 0.9277 7.23%
No. 1 1 0.9339 6.61%
Case 8
No. 2 1 0.8636 13.64%
No. 1 1 0.8866 11.34%
Case 9
No. 2 1 0.9246 7.54%
Table 8. The depth and angle evaluations of Case 10 and Case 11.
Actual Reconstructed Estimation Error Actual Incidence Reconstructed Estimation Error

Specimen
Depth (mm) Depth (mm) of Depth Angle (◦ ) Angle (◦ ) of Angle
Case 10 1 0.8510 14.90% 90 92.37 2.63%
Case 11 1 0.8345 16.55% 135 143.46 6.27%
5. Discussion
In this paper, the size of a spatial window is 16 mm, while the diameter of a circular flat bottom
hole is no more than 10 mm. Hence, this spatial window will not only contain the propagating
waves located at the damage, but also contain those located at the undamaged area. According to the
dispersion curves, it should be noted that the wavenumber of the A0 mode will decrease with the
increase in the frequency thickness product. More specifically, the wavenumber located at the damage
is larger than that located at the undamaged area. Moreover, the deeper the circular flat bottom hole is,
the more prominent the difference is. Therefore, this window effect cannot be ignored.
(1) The depth estimation of the deep hole is more accurate than that of the shallow one, as revealed
by comparing the results of Case 1 and Case 2.
The frequency-thickness product of the shallow hole is much larger than that of the deep hole,
which means that the wavenumber of the former is smaller than that of the latter. Additionally,
there is no significant difference between the shallow hole and the undamaged area in terms of the
wavenumber. According to Equation (11), it is easy to understand that the shallow hole will be easily
affected by the window effect relative to the deep hole.
(2) The depth estimation for the big hole is more accurate than that of the small one.
This conclusion was drawn by comparing the results of Case 1 and Case 3.
One reasonable explanation for this is that the local wavenumber is affected by the difference
in size between the defect and the spatial window. The small diameter of a circular flat bottom hole
means that the spatial window will contain more propagating waves located at the undamaged area
where the wavenumbers are small. According to Equation (11), it is easy to understand that the local
wavenumber will decrease as the lower wavenumbers are present in the spatial window. The smaller
the diameter of a circular flat bottom hole is, the smaller the local wavenumber and the reconstructed
depth will be.
(3) The results of Case 4 and Case 5 indicate that the better test result appears under the big gap
between defects when two circular flat bottom holes are symmetrical to the excitation source.
Apparently, multi-reflection between two defects is low when they are far apart from each other.
Furthermore, they are affected at the same level as they are symmetrical to the excitation source.
(4) The circular flat bottom hole far away from the excitation source seems smaller than the near
one when the excitation source and two identical defects are in the same line. Moreover, the closer the
two defects are, the more obvious the effect is. This conclusion was obtained by comparing the results
of Case 6 and Case 7.
179
Appl. Sci. 2018, 8, 1600
In order to determine the explanation for the above phenomenon, several numerical simulations
were conducted under different conditions, which are listed in Table 9. The wavenumber distributions
of the spatial window are shown in Figure 23. There is only one defect existing in Figure 23a–c, while
there are two defects coexisting in Figure 23d–f. The wavenumber distributions of Figure 23a,d are
obtained under the spatial window whose center is at (200, 210). Great differences between Figure 23a,d
can be found. The wavenumber distributions of Figure 23b,e are obtained under the spatial window
whose center is at (200, 215). There are some differences between Figure 23b,e. The wavenumber
distributions of Figure 23c,f are obtained under the spatial window whose center is at (200, 220).
Only a few differences between Figure 23c,f can be found. The fundamental reason for this is that the
defect near to the excitation source can change the incident wave in direction and amplitude, which
dominates the distribution of the wavenumber. Furthermore, it can be clearly found that the impact
which the near defect exerts on the far one is gradually reduced with the increase of the distance
between them when the excitation source and two identical defects are in the same line.
Figure 23. The wavenumber distributions of the spatial window when different conditions are taken
into consideration. (a) Condition 1; (b) condition 2; (c) condition 3; (d) condition 4; (e) condition 5;
(f) condition 6. The conditions are listed in Table 9.
Table 9. The conditions of Figure 23.
Excitation Central Damage Damage Central Position Spatial

Condition Damage Type Source Position Position of Radius Depth of Spatial Window
(mm) Damage (mm) (mm) (mm) Window (mm) Radius (mm)
Condition 1 One flat bottom hole (200, 100) (200, 210) 5 1 (200, 210) 8
No. 1: (200, 210) 5 1
Condition 4 Two flat bottom holes (200, 100) (200, 210) 8
No. 2: (200, 190) 5 1
No. 1: (200, 215) 5 1
No. 2: (200, 185) 5 1
No. 1: (200, 220) 5 1
No. 2: (200, 180) 5 1
180
Appl. Sci. 2018, 8, 1600
(5) The results of Case 8 and Case 9 indicate that the impact which the big hole exerts on the small
one is related to their positions when the excitation source and multiple defects are in the same line.
The direction and amplitude of the incident wave can be changed obviously when the big hole is
in front of the small one. Hence, the wavenumber distributions located at the small hole also change.
The impact that the big hole exerts on the small one comes mainly from the reflection wave when the
big hole is behind the small one. However, it is the incident wave that plays a crucial role in changing
the wavenumber distributions, because the incident wave has a larger amplitude than the reflection
wave. Therefore, the test result for the small hole will only change a little when the big hole is behind it.
(6) The depth errors for Case 10 and Case 11 are more than 10%.
The length of the rectangular groove is larger than that of the spatial window. However, the
width of the former is still far smaller than that of the latter. Hence, the local wavenumber is seriously
affected by the window effect.
(7) Several measures are adopted to improve the test quality when the proposed approach is
applied for an unknown damage.
In practice, both the damage size and its position are usually unknown. A large amount of data
will be acquired if the entire structure is inspected. Thus, a preliminary step is desired. More specifically,
it is easy to acquire the approximate location of a defect by using reliable methods, such as DAS or
the compact phased array (CPA) method [40]. Then a small scanning region where the defect is likely
to exist can be obtained. Under the circumstance, the local wavenumber approach can be conducted
without consuming much time. The spatial window size is at least twice the wavelength of the excited
mode. It should be noted that the best test result is not always obtained under the minimum size of
the spatial window. In fact, the best test result is obtained under the optimum size due to the fact
that the optimum size contributes to diminishing the spatial window effect. The optimum size is also
fixed when the central frequency of the excitation signal and the excited mode are fixed. The optimum
size can be attained by adjusting the size of the spatial window until the maximum wavenumber of
the wavenumber map is acquired. Therefore, patience is important during the period of seeking the
optimum size. The process of improving the test quality is shown as below.
The excitation signal with a higher central frequency is adopted as the wavelength of the excited
mode is small. Under this condition, the minimum size of the spatial window may be smaller than
the optimum size. The wavelength of the A0 mode is about 5 mm when the central frequency of the
excitation signal is 400 kHz. Hence, the minimum radius of the spatial window is 5 mm. The radius
will be increased at an interval of 1 mm. The test results obtained under various window radii are
shown in Figure 24. The maximum wavenumber obtained under various window radii is shown in
Figure 25.
Figure 24 shows that the identified damage region will expand with the increase of the window
radius. It was found that the test results are poor when the radius is small. The identified damage
region is almost equal to the actual size under a special window radius (8 mm). Figure 25 shows that
the maximum wavenumber increases with the size of the spatial window and gradually approaches
to the theoretical value at the early stage. The maximum wavenumber is very close to the theoretical
value under a special window radius (8 mm). However, the maximum wavenumber will decrease
when the size of the spatial window continues to be increased.
It is worth that the wavelength of the excited mode is large when the central frequency of the
excitation signal is low. Hence, it may lead to the minimum size of the spatial window being larger
than the optimum size. Under this circumstance, a good test result is obtained at the minimum size of
a spatial window. The test results are shown in Figures 9 and 10.
181
Appl. Sci. 2018, 8, 1600
Figure 24. The test results obtained under various window radii. (a) 5 mm; (b) 6 mm; (c) 7 mm;
(d) 8 mm; (e) 9 mm; (f) 10 mm; (g) 11 mm; (h) 12 mm. The white circle represents the actual location of
the damage. The central frequency of the excitation signal is 400 kHz. A circular flat bottom hole is
located in (185, 185).
(8) The limitation of the proposed approach should be taken into consideration when it is applied
for characterizing defects existing in anisotropic materials.
The delamination of composite laminated plates was taken as an example to illustrate the proposed
approach. The interactions between Lamb waves and the delamination are complex. It is difficult to
identify and isolate a specific mode for the reconstruction analysis. Therefore, the average wavenumber
is less accurate. In addition, the position of the delamination cannot be quantified precisely.
In summary, the test quality can be affected by the spatial window effect and the reflection
between multiple defects. It is necessary to avoid making the excitation source and multiple defects
182
Appl. Sci. 2018, 8, 1600
in the same line in order to reduce the interference between adjacent defects. The optimum size of
a spatial window need be sought based on the test results. Increasing the central frequency of the
excitation signal is a way to decrease the wavelength of the excited mode and helps to improve the test
quality. However, the high frequency can easily produce complex modes. Therefore, it is important to
maintain a balance between increasing the central frequency of the excitation signal and reducing the
mode complexity.
Figure 25. The maximum wavenumber obtained under various window radii. The central frequency
of the excitation signal is 400 kHz.
6. Future Work
Given that the local wavenumber approach has many advantages in structural health monitoring
(SHM), we will apply this approach for damage detection in adhesive-bonded concrete structures.
The main work will be focused on identifying the local debonding and the weakened adhesion.
A non-contact system will be composed of two low frequency air coupled transducers. One will be
located at a fixed place for actuating a desired guided wave mode by setting an appropriate incidence
angle. The other will be used for receiving the signals with the same inclined angle and for scanning
over the target area with a spatial interval. The work will be full of challenges as separating the desired
mode from multi-modes is not easy and the attenuation existing in the concrete structure may lead to
a low signal-to-noise ratio. The detailed solutions will be introduced in the future work.
7. Conclusions
The Lamb wave local wavenumber approach for characterizing flat bottom defects (including
two circular flat bottom holes and a rectangular groove) in an isotropic thin plate was presented in this
paper. The hybrid and non-contact system is composed of an air-coupled transducer (ACT) and a laser
Doppler vibrometer (LDV). On the one hand, the A0 mode is easily actuated by the ACT with a special
incidence angle. On the other hand, the out-of-plane velocity over the scanning area can be recorded
by an LDV.
Both the simulation results and the experimental results demonstrated that the damage can be
characterized in shape, location, and size. Compared with other methods, the proposed approach
can be used to reconstruct the thickness distribution over the scanning area when the wavenumber
is combined with the phase velocity dispersion curve. Consequently, the depth of the flat bottom
defect also can be quantified. In addition, adjacent defects are clearly characterized and the direction
of the inclined rectangular groove can be obtained by calculating the image moments under grayscale.
It must be noted that an appropriate window size contributes to improving the test quality.
Author Contributions: G.F. wrote this article and accomplished the FEM analysis; H.Z. (Haiyan Zhang) was the
principle researcher on this project and provided many valuable suggestions. H.Z. (Hui Zhang) designed the
three-dimensional finite element model and performed the simulation; W.Z. established the experimental system
and X.C. analyzed the experimental data.
183
Appl. Sci. 2018, 8, 1600
Acknowledgments: The work in this paper is supported by the National Natural Science Foundation of China
(Grant Nos. 11674214, 11874255, 11474195, and 51478258), and the Key Technology R&D Project of Shanghai
Committee of Science and Technology (Grant No. 16030501400).
References
1. Park, I.; Jun, Y.; Lee, U. Lamb wave mode decomposition for structural health monitoring. Wave Motion 2014,
51, 335–347. [CrossRef]
2. Wang, W.; Bao, Y.; Zhou, W.; Li, H. Sparse representation for Lamb-wave-based damage detection using a
dictionary algorithm. Ultrasonics 2018, 87, 48–58. [CrossRef] [PubMed]
3. Lu, G.; Li, Y.; Wang, T.; Xiao, H.; Huo, L.; Song, G. A multi-delay-and-sum imaging algorithm for damage
detection using piezoceramic transducers. J. Intell. Mater. Syst. Struct. 2017, 28, 1150–1159. [CrossRef]
4. Khodaei, Z.S.; Aliabadi, M.H. Assessment of delay-and-sum algorithms for damage detection in aluminum
and composite plates. Smart Mater. Struct. 2014, 23, 628–634.
5. Zhang, Y.; Li, D.; Zhou, Z. Time Reversal Method for Guided Waves with Multimode and Multipath on
Corrosion Defect Detection in Wire. Appl. Sci. 2017, 7, 424. [CrossRef]
6. Zhang, H.Y.; Cao, Y.P.; Sun, X.L.; Chen, X.H.; Yu, J.B. A time reversal damage imaging method for structure
health monitoring using Lamb waves. Chin. Phys. B 2010, 19, 448–455. [CrossRef]
7. Zhang, H.Y.; Ruan, M.; Zhu, W.F.; Chai, X.D. Quantitative damage imaging using Lamb wave diffraction
tomography. Chin. Phys. B 2016, 25, 25–31. [CrossRef]
8. Ohara, Y.; Takahashi, K.; Ino, Y.; Yamanaka, K.; Tsuji, T.; Mihara, T. High-selectivity imaging of closed
cracks in a coarse-grained stainless steel by nonlinear ultrasonic phased array. NDT E Int. 2017, 91, 139–147.
[CrossRef]
9. Ambrozinski, L.; Piwakowski, B.; Stepinski, T.; Uhl, T. Evaluation of dispersion characteristics of multimodal
guided waves using slant stack transform. NDT E Int. 2014, 68, 88–97. [CrossRef]
10. Sohn, H.; Kim, S.B. Development of dual PZT transducers for reference-free crack detection in thin plate
structures. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2010, 57, 229–240. [CrossRef] [PubMed]
11. Marks, R.; Clarke, A.; Featherston, C.; Paget, C.; Pullin, R. Lamb Wave Interaction with Adhesively Bonded
Stiffeners and Disbonds Using 3D Vibrometry. Appl. Sci. 2016, 6, 12. [CrossRef]
12. Xu, K.; Ta, D.; Hu, B.; Laugier, P.; Wang, W. Wideband Dispersion Reversal of Lamb Waves. IEEE Trans.
13. Ayers, J.; Apetre, N.; Ruzzene, M.; Sabra, K. Measurement of Lamb wave polarization using a
one-dimensional scanning laser vibrometer (L). J. Acoust. Soc. Am. 2011, 129, 585–588. [CrossRef] [PubMed]
14. Wu, T.Y.; Ume, I.C. Fundamental study of laser generation of narrowband Lamb waves using superimposed
line sources technique. NDT E Int. 2011, 44, 315–323. [CrossRef]
15. Xu, K.; Ta, D.; Moilanen, P.; Wang, W. Mode separation of Lamb waves based on dispersion compensation
method. J. Acoust. Soc. Am. 2012, 131, 2714–2722. [CrossRef] [PubMed]
16. Tian, Z.; Yu, L. Lamb wave frequency–wavenumber analysis and decomposition. J. Intell. Mater. Syst. Struct.
2014, 25, 1107–1123. [CrossRef]
17. Yu, L.; Tian, Z. Lamb wave Structural Health Monitoring Using a Hybrid PZT-Laser Vibrometer Approach.
Struct. Health Monit. 2013, 12, 469–483. [CrossRef]
18. Yu, L.; Leckey, C.A.C.; Tian, Z. Study on crack scattering in aluminum plates with lamb wave
frequency-wavenumber analysis. Smart Mater. Struct. 2013, 22, 1–12. [CrossRef]
19. Yu, L.; Tian, Z.; Leckey, C.A.C. Crack imaging and quantification in aluminum plates with guided wave
wavenumber analysis methods. Ultrasonics 2015, 62, 203–212. [CrossRef] [PubMed]
20. Rogge, M.D.; Leckey, C.A.C. Characterization of impact damage in composite laminates using guided
wavefield imaging and local wavenumber domain analysis. Ultrasonics 2013, 53, 1217–1226. [CrossRef]
[PubMed]
21. Rao, J.; Ratassepp, M.; Fan, Z. Quantification of thickness loss in a liquid-loaded plate using ultrasonic
guided wave tomography. Smart Mater. Struct. 2017, 26. [CrossRef]
22. Rao, J.; Ratassepp, M.; Fan, Z. Guided Wave Tomography Based on Full Waveform Inversion. IEEE Trans.
184
Appl. Sci. 2018, 8, 1600
23. Chronopoulosa, D.; Droz, C.; Apalowo, R.; Ichchou, M.; Yan, W.J. Accurate structural identification for
layered composite structures, through a wave and finite element scheme. Compos. Struct. 2017, 182, 566–578.
[CrossRef]
24. Li, H.; Zhou, Z. Air-Coupled Ultrasonic Signal Processing Method for Detection of Lamination Defects in
Molded Composites. J. Nondestruct. Eval. 2017, 36, 1–13. [CrossRef]
25. Sunarsa, T.Y.; Aryan, P.; Jeon, I.; Park, B.; Liu, P.; Sohn, H. A Reference-Free and Non-Contact Method for
Detecting and Imaging Damage in Adhesive-Bonded Structures Using Air-Coupled Ultrasonic Transducers.
Materials 2017, 10, 1402. [CrossRef] [PubMed]
26. Jeon, J.Y.; Gang, S.; Park, G.; Flynn, E.; Kang, T.; Han, S.W. Damage detection on composite structures with
standing wave excitation and wavenumber analysis. Adv. Compos. Mater. 2017, 26, 53–65. [CrossRef]
27. Harb, M.S.; Yuan, F.G. A rapid, fully non-contact, hybrid system for generating Lamb wave dispersion
curves. Ultrasonics 2015, 61, 62–70. [CrossRef] [PubMed]
28. Urban, M.W.; Nenadic, I.Z.; Qiang, B.; Bernal, M.; Chen, S.; Greenleaf, J.F. Characterization of material
properties of soft solid thin layers with acoustic radiation force and wave propagation. J. Acoust. Soc. Am.
2015, 138, 2499–2507. [CrossRef] [PubMed]
29. Lee, C.; Park, S. Flaw Imaging Technique for Plate-Like Structures Using Scanning Laser Source Actuation.
Shock Vib. 2014, 9, 1–14. [CrossRef]
30. Tian, Z.; Yu, L.; Leckey, C.A.C. Delamination detection and quantification on laminated composite structures
with Lamb waves and wavenumber analysis. J. Intell. Mater. Syst. Struct. 2015, 26, 1723–1738. [CrossRef]
31. Juarez, P.D.; Leckey, C.A.C. Multi-frequency local wavenumber analysis and ply correlation of delamination
damage. Ultrasonics 2015, 62, 56–65. [CrossRef] [PubMed]
32. Nabili, M.; Geist, C.; Zderic, V. Thermal safety of ultrasound-enhanced ocular drug delivery: A modeling
study. Med. Phys. 2015, 42, 5604–5615. [CrossRef] [PubMed]
33. Bhuiyan, M.Y.; Shen, Y.; Giurgiutiu, V. Interaction of Lamb waves with rivet hole cracks from multiple
directions. J. Mech. Eng. Sci. 2016, 231, 2974–2987. [CrossRef]
34. Bhuiyan, M.Y.; Shen, Y.; Giurgiutiu, V. Guided Wave Based Crack Detection in the Rivet Hole Using Global
Analytical with Local FEM Approach. Materials 2016, 9, 602. [CrossRef] [PubMed]
35. Testoni, N.; Marchi, L.D.; Marzani, A. Detection and characterization of delaminations in composite plates
via air-coupled probes and warped-domain filtering. Compos. Struct. 2016, 153, 773–781. [CrossRef]
36. Xiao, B.; Li, L.; Li, Y.; Li, W.; Wang, G. Image analysis by fractional-order orthogonal moments. Inf. Sci. 2017,
382, 135–149. [CrossRef]
37. Martín, J.A.H.; Santos, M.; Lope, J.D. Orthogonal variant moments features in image analysis. Inf. Sci. 2010,
180, 846–860. [CrossRef]
38. Zhou, C.; Zhang, B.; Lin, K.; Xu, D.; Chen, C.; Yang, X.; Sun, C. Near-infrared imaging to quantify the feeding
behavior of fish in aquaculture. Comput. Electron. Agric. 2017, 135, 233–241. [CrossRef]
39. Losada, R.A.; Pellisier, V. Designing IIR Filters with a Given 3-dB Point. IEEE Signal Process. Mag. 2005, 22,
95–98. [CrossRef]
40. Senyurek, V.Y.; Baghalian, A.; Tashakori, S.; McDaniel, D.; Tansel, I.N. Localization of multiple defects using
the compact phased array (CPA) method. J. Sound Vib. 2018, 413, 383–394. [CrossRef]
185
applied
sciences
Article
Ultrasonic Guided Wave-Based Circumferential
Scanning of Plates Using a Synthetic Aperture
Focusing Technique
Jianjun Wu 1 , Zhifeng Tang 2, *, Keji Yang 1 , Shiwei Wu 1 and Fuzai Lv 1
1 State Key Laboratory of Fluid Power and Mechatronic Systems, School of Mechanical Engineering,
Zhejiang University, 38 Zheda Road, Hangzhou 310027, China; 11425023@zju.edu.cn (J.W.);
yangkj@zju.edu.cn (K.Y.); 06jxlwsw@zju.edu.cn (S.W.); lfzlfz@zju.edu.cn (F.L.)
2 Institute of Advanced Digital Technologies and Instrumentation, School of Biomedical Engineering and
Instrument Science, Zhejiang University, 38 Zheda Road, Hangzhou 310027, China
* Correspondence: tangzhifeng@zju.edu.cn; Tel.: +86-138-5713-1010
Received: 11 July 2018; Accepted: 6 August 2018; Published: 7 August 2018
Featured Application: Ultrasonic guided wave scanning of plates with the transducer moving
along a circumferential path outside the tested area.
Abstract: Tanks are essential facilities for oil and chemical storage and transportation. As indispensable
parts, the tank floors have stringent nondestructive testing requirements owing to their severe
operating conditions. In this article, a synthetic aperture focusing technology method is proposed
for the circumferential scanning of the tank floor from the edge outside the tank using ultrasonic
guided waves. The zeroth shear horizontal (SH0) mode is selected as an ideal candidate for plate
inspection, and the magnetostrictive sandwich transducer (MST) is designed and manufactured
for the generation and receiving of the SH0 mode. Based on the exploding reflector model (ERM),
the relationships between guided wave fields at different radii of polar coordinates are derived in
the frequency domain. The defect spot is focused when the sound field is calculated at the radius of
the defect. Numerical and experimental validations are performed for the defect inspection in an
iron plate. The angular bandwidth of the defect spot is used as an index for the angular resolution.
The results of the proposed method show significant improvement compared to those obtained by
the B-scan method, and it is found to be superior to the conventional method—named delay and
sum (DAS)—in both angular resolution and calculation efficiency.
Keywords: SH0 mode; circumferential scanning; synthetic aperture focusing; exploding

reflector model
1. Introduction
Tanks are essential facilities for the storage and transportation of petroleum products.
Thousands of tanks may be in service at the same time in one oil depot. Because of the harsh
working conditions and corrosive storage materials, various kinds of defects, which are hidden
hazards, may occur in the tanks. The structural integrity failure of the tanks may cause catastrophic
accidents such as environmental pollution and explosions [1]. Thus, tanks require stringent inspection
for their long-term and high-level safety. Tank floors are indispensable parts of the tanks. It is found
that 80% of corrosion, which makes up the majority of the defects in the tanks, occur in the tank floors,
making them key areas in the overhauls [2]. The most common non-destructive testing technologies
for tank floors based on flux leakage [3–5] and ultrasonics [6] are limited for the low working efficiency
of testing “step by step.” The tanks must stop their service and be emptied before the inspection.
The cost and time for the overhaul of the tank floors can amount to hundreds of thousands of dollars

Appl. Sci. 2018, 8, 1315
and dozens of days. Recently, a technology based on sound emission has been introduced for the
structural health monitoring of tanks in service [7–9]. Elastic waves caused by the crack growth and
liquid leakage are collected and analyzed for the damage detection. However, useful information is
difficult to extract from the signals because of the low signal-to-noise ratios, and the localization of
the damage is dissatisfactory with a limited number of transducers. Therefore, it is urgent and highly
desirable to develop new methods for testing tank floors in service.
Ultrasonic guided wave methods have been proven to have tremendous potential in the rapid
inspections of structures with long and slender dimensions such as pipes [10,11], plates [12–14], and
rails [15]. For the inspection of plate structures such as tank floors, large areas can be inspected
at the same time with one transducer using guided waves. Structural health monitoring based on
guided waves has been proposed for years. The transducers are permanently installed on the plate
around the monitored region. The transducers work as actuators, in turn, and all the others are
sensors. The received signals are compared with those acquired when the integrity of the tank floors
was confirmed. The coefficients to evaluate the signal changes can be amplitude differences of the
direct waves [16] and the variations of the whole signals, named the signal difference coefficient
(SDC) [17–21]. Different imaging methods are introduced for guided wave monitoring, such as the
filtered back-projection method (FBP) [22], the interpolated FBP (IFBP) [23], the algebraic reconstruction
technique (ART) [24], and the probabilistic reconstruction algorithm (PRA) [25], which are all basically
drawn from computational tomography (CT). Recently, a method called full waveform inversion
(FWI) has been introduced for guided wave monitoring [26,27]. Guided waves with high frequency
dispersions are used, and the model of the monitored area is constructed and revised in the frequency
domain to match the signals acquired by the model and those collected in the tests. This method
performs well in corrosion imaging. However, it should be mentioned that numerous transducers are
needed for high sensitivities and localization accuracies in guided wave monitoring, and multiple
iterations are required in these methods for high resolutions, which may be time-consuming for large
amounts of data.
Guided wave scanning is an alternative method for plate inspection. As shown in Figure 1,
the tank floor in service can be evaluated with one transducer moving along the edge outside the
tank [28]. The spots of the defects in the image of the original signals (i.e., the B-scan image in this
study) have long lateral trailing because of the expansion of the sound fields in the plates. To improve
the lateral resolution, the synthetic aperture focusing technique (SAFT) is developed. The traditional
method of SAFT is called delay and sum (DAS) [29,30]. The wave packets of the defect in different
signals can be superposed in the image with the distances between the defect and the transducer
locations in the scanning. This method is theoretically applicable for arbitrary scanning paths of
the transducer. However, the method is carried out in the time domain with poor computational
efficiency, and residual trailing exists in the image. Another method of SAFT is proposed based on
sound field migration in the frequency domain, providing better computational efficiency as well as
imaging quality compared with the DAS method [31–33]. However, the sound field migration of this
method is deduced in the Cartesian coordinate system, which can only manage signals acquired by
the transducer moving along a straight path. The paths of the transducer for the scanning of the tank
floors are circular, making this method inapplicable. Efficient methods of SAFT for guided wave-based
circumferential scanning have not been seen yet.
In this paper, a method of SAFT for circumferential scanning (CSAFT) is proposed. In Section 2,
the shear horizontal mode in the plates, named SH0, is selected for inspection, and the sound field of
the magnetostrictive sandwich transducer (MST) used in this study is calculated; the imaging method
is developed based on sound field migration in the polar coordinate systems. In Section 3, the proposed
method is verified by numerical simulations and compared with other two methods, the B-scan and
DAS methods, for different affecting factors. Experimental verifications are carried out in an iron plate
in Section 4 and conclusions follow in Section 5.
188
Appl. Sci. 2018, 8, 1315
‫͵ݑ‬
‫ͳݑ‬
ܱͳ
‫ʹݑ‬ ʹ
ͳ
Ƚͳ Ƚʹ ܱʹ
Figure 1. The diagram of the circumferential scanning using guided waves.
2. Method
2.1. Shear Horizontal Waves

As shown in Figure 1, in circumferential scanning using guided waves, the transducer to generate
and receive guided waves moves around the origin point O2 at the radius R1 . The wave front of the
incident guided waves expands during the propagation. The reflection waves are generated if the
defects are covered by the wave front of the incident waves. The flight time and peak values of the
reflection waves are used to evaluate the location and size of the defects, respectively.
Shear horizontal waves (SH waves) and Lamb waves are two kinds of mechanical waves
propagating in a solid plate. For waves propagating along the direction u1 as in Figure 1, displacements
along the direction u3 occur in Lamb waves, which are perpendicular to the plane of the plate, but the
displacements of SH waves are along the direction u2 within the plane of the plate. Considering that
the tank floors in service are in contact with liquids (storage materials, such as different kinds of oil)
and that the attenuation of shear waves in liquids is much higher than that of longitudinal waves,
SH waves may be better candidates for an inspection with less energy leakage. Researches of torsional
waves in pipes filled with liquids and bars immersed in liquids can be found in [34,35], respectively,
as references. Quantitative comparisons of the SH and Lamb waves will be carried out further under
the conditions of different kinds of petroleum products with various volumes. And affecting factors
such as the curvatures of the tank floors caused by the gravities of the stored materials and sludge
deposition under the tank floors will be investigated as well. The SH waves are hardly affected
by liquids [36]. Therefore, in this study, with the purpose of verification of the proposed imaging
method, the model is simplified as an isotropic plate with the thickness d without contact of liquids,
the dispersion equation of SH waves can be expressed as [37]
ω2 ω2 nπ 2
2
− 2 = (1)
cT cp d
"
E
cT = (2)
2(1 + μ ) ρ
where ω = 2π f , f , c p , and n are frequency, phase velocity, and order, respectively, and c T is the
fundamental shear wave velocity, which is determined by mechanical properties of the plate such as
density ρ, Young’s modulus E, and Poisson’s ratio μ. Combining the definition of the wave number
k = cωp , the group velocity c g = dω
dk , and Equations (1) and (2), dispersion curves of phase velocity and
group velocity can be calculated:
fd
c p ( f d) = 2c T (3)
(2 f d)2 − (nc T )2
189
Appl. Sci. 2018, 8, 1315
#
$
$ (n/2)2
c g ( f d ) = c T %1 − (4)
( f d/c T )2
where f d is the frequency-thickness product. Dispersion curves of phase velocity and group velocity
for the iron plates with mechanical properties ρ = 7850 kg/m3 , E = 210 GPa, and μ = 0.3 are shown
in Figure 2. All the modes are named by their orders, and for the wave mode SHn, a frequency lower
limit exists when n is no less than 1, which is called the cutoff frequency.
nc T
( f d)n = (5)
2
The thickness of the tank floor commonly used is between 6 and 19 mm [38]. In this study,
the thickness d of the iron plate used for the verification in numerical simulations and experiments
is selected as 4 mm for the easy operations in the laboratory. The wave structures of SH0, SH1, SH2,
and SH3 modes in the direction u2 are shown in Figure 3. It can be observed that the pure SH0 mode
without dispersion can be generated when the excitation frequency is lower than the cutoff frequency
of the SH1 mode. Compared with high order modes (SH1, SH2 and SH3), the wave structure of the
SH0 mode is uniform along the thickness, giving the same detectability for the defects within the plates
and on the surfaces. With these advantages, the SH0 mode is selected for inspection in this study.
(a) (b)
Figure 2. Dispersion curves of iron plates: (a) phase velocity curves; (b) group velocity curves.
(a) (b)
(c) (d)
Figure 3. Wave structures of shear horizontal (SH) modes in the main vibration direction u2 : (a) SH0;
(b) SH1; (c) SH2; (d) SH3.
190
Appl. Sci. 2018, 8, 1315
2.2. The Sound Field of the Transducer

Piezoelectric and magnetostrictive transducers are commonly used in guided wave generation.
Because of the differences in the principles of energy conversion and transducer structures, piezoelectric
transducers are good at out-of-plane displacement generation, while magnetostrictive transducers
have advantages in in-plane displacement generation. The magnetostrictive effect of the ferromagnetic
materials is made use of in the wave generation, and the inverse magnetostrictive effect works in
receiving waves. Details of the working principles of this kind of transducer can be found in [39].
The transducer used in this study is named the magnetostrictive sandwich transducer (MST),
with a structure shown in Figure 4. The magnetostrictive strip is the main component for energy
conversion in both wave generation and receiving, with the residual magnetism along the direction
u2 by pre-magnetization. The soft magnet is used to reduce the magnetic resistance of the dynamic
magnetic fields, improving energy conversion efficiency. There are two coils in the MST to generate
and sense the dynamic magnetic fields. Coil 1 is λ/4 ahead of Coil 2 in the direction u1 . λ is the
wavelength of the guided wave. The burst signals in Coil 1 and Coil 2 have a phase difference of π/2.
Wires in the working region of the MST are distributed with equal intervals and numbered with wn
from 1 to 40. The current direction Cd and initial phase Cip of all the wires are listed in Table 1.
Table 1. The current direction Cd and initial phase Cip for each wire.
Name Value
wn 1~5 6~10 11~15 16~20 21~25 26~30 31~35 36~40
Cd (wn) −1 −1 1 1 22121 −1 1 1
Cip (wn) π/2 0 π/2 0 π/2 0 π/2 0
Each wire in the working region can be regarded as a line sound source, and the sound field Swn
for the wnth wire can be derived as [40]
& ⎡ ⎤

2 ⎣ sin 2 kL sin(θ ) ⎦ j(kr −ωt+ π )
1

Swn r , θ , t ≈ Sm L e 4 (6)
πkr
2 kL sin( θ )
1
where the wave number k = 2π

λ , Sm is a constant coefficient, and r and θ denote the relative position
between the wnth wire and the receiving point as in Figure 4. L is half the length of the working region.
Equation (6) is valid when r L. The sound field of the MST can be calculated as the sound field
combinations of all wires:
S ≈ ∑wn=1 Cd (wn)Swn e jCip (wn) .
40
(7)
As mentioned in Section 2.1, the thickness of the iron plate used in this study is 4 mm. The group
velocity c g and phase velocity c p of the SH0 mode are both 3208 m/s, as in Figure 2. The wave length
λ of the SH0 mode is 25.1 mm as the width of the magnetostrictive strip available is 50.2 mm. Thus,
the main frequency f of the designed MST is 128 kHz, which is below the cutoff frequency of the SH1
mode (401 kHz). It should be noted that, as mentioned above, the main frequency of the transducer
should be determined with the consideration of both detection sensitivities and testing ranges in field
tests. The normalized sound fields of the MSTs with different lengths are shown in Figure 5. The waves
are controlled to propagate in one direction, and the amplitudes of the main lobes in the sound fields
are much larger than those of the side lobes. It can also be seen that the widths of the main lobes
decrease with the increase in the lengths of the MSTs. The divergence angle θd , defined as the largest
angle at which the amplitudes are no less than half of the largest amplitude, is used to denote the
width of the main lobes. The relationship between the divergence angles and the lengths of the MSTs
from 2 mm to 150 mm are shown in Figure 6.
191
Appl. Sci. 2018, 8, 1315
‫ͳݑ‬
‫ݎ‬Ԣ ‫ݎ‬
ܱͳ ‫ʹݑ‬
Figure 4. The structure of the magnetostrictive sandwich transducer (MST).
‫ͳݑ‬ ‫ͳݑ‬
‫ʹݑ‬ ‫ʹݑ‬
ߠ݀ ൌ ͺ͹ι ߠ݀ ൌ ͸ͻι
(a) (b)
‫ͳݑ‬ ‫ͳݑ‬
‫ʹݑ‬ ‫ʹݑ‬
ߠ݀ ൌ ͷͳι ߠ݀ ൌ ͶͲι
(c) (d)
Figure 5. The sound fields and divergence angles of the MSTs with different lengths: (a) 2L = 30 mm;
(b) 2L = 50 mm; (c) 2L = 70 mm; (d) 2L = 90 mm.
192
Appl. Sci. 2018, 8, 1315
ș d(deg)
Figure 6. The relationship between the divergence angles of the main lobes and the lengths of the MSTs.
2.3. Synthetic Focusing Imaging for Circumferential Scanning

The whole procedure of the inspection with the transducer at one location can be described
as follows:
Step 1: The transducer generates guided waves towards the defect;

Step 2: The defect is motivated and generates scattering waves;
Step 3: Parts of the scattering waves are received by the transducer.
When the defect is motivated by the incident pure SH0 waves, other modes may exist in the
scatter waves because of the mode conversions. These modes are not dominant and they can hardly
be detected by the MST in the experiments. Therefore, the mode conversion is not considered in this
study. Because of the reciprocity of the sound field and the constant group velocity c g of the SH0 mode,
Steps 1 and 3 can be described in the same way. The defect can be regarded as a sound source that
generates guided waves at time t = 0; then, the guided waves propagate to the transducer. Since the
sonic path in this assumption is half of the actual path, the group velocity ĉ g is taken as half of the
actual value c g to maintain the flight time of the defect, i.e., ĉ g = c g /2. The above equivalent model
is also called the exploding reflector model (ERM) [41]. As in Figure 1, the guided waves generated
by the defect at the radius R2 expand before being received by the transducer moving at the radius
R1 . If R1 is equal to R2 , there will be no expansion of the guided waves, which are simply generated
from the defect at the time t = 0. An optimal lateral resolution can be obtained from the guided wave

fields at the radius R2 with the time term e−iωt t = 0 = 1. The relationships of the guided wave fields
at different radii can be deduced as follows.
In Figure 1, the polar coordinate O2 is constructed without considering the direction along the
thickness, in which the displacement of the SH0 mode is uniform. The wave equation in a polar
coordinate can be written as [42]:
!
∂2 1 ∂ 1 ∂2 1 ∂2
+ + − p( R, α, t) = 0 (8)
∂R2 R ∂R R2 ∂α2 ĉ g ∂t2
where p( R, α, t) is the wave field of the point ( R, α) at time t, which can be written in the form of an
inverse Fourier transform:
'
p( R, α, t) = P( R, k α , ω )e j(kα α+ ωt) dk α dω (9)
where k α and ω are the angular wave number and angular frequency, respectively. P( R, k α , ω ) is the
two-dimensional spectrum of p( R, α, t), which can be solved by substituting Equations (9) into (8):
(1) (2)
P( R, k α , ω ) = Akα Hkα (k R R) + Bkα Hkα (k R R) (10)
193
Appl. Sci. 2018, 8, 1315
(1) (2)
where Akα Hkα (k R R) and Bkα Hkα (k R R) denote the guided wave propagating away from the original
( p)
point and towards the original point, respectively. Hkα (p = 1, 2) is the pth kind of Hank function of
the order k α k R is the wave number along the radius.
⎧ & &
⎪
⎪ ω2
− 2, ω2
− kα 2 ≥ 0
⎨ ĉ g 2
k α ĉ g 2
kR = & . (11)
⎪
⎪ ω2
⎩ 0, ĉ 2
− kα 2 < 0
g
The guided waves generated by the defect propagate away from the original point to the
transducer; thus, the transfer function for the guided wave fields at different radii can be derived as
(1)
Hkα (k R ( R − ΔR))
P( R − ΔR, k α , ω ) = P( R, k α , ω ) (1)
(12)
Hkα (k R R)
Equation (12) can be simplified using the Rayleigh–Sommerfeld diffraction formula [43]:
& &
ω2 − kα 2
R − ΔR −iΔR ĉ g 2 R( R−ΔR)
P( R − ΔR, k α , ω ) ≈ P( R, k α , ω ) e (13)
R
( α − α )2
This function is valid if cos(α1 − α2 ) is approximately equal to 1 − 1 2 2 With a
given step ΔR, the two-dimensional spectrums of the guided wave fields at the radii
( R − ΔR, R − 2ΔR, R − 3ΔR, . . . , R − mΔR) can be derived with those collected by the transducer
at the radius R with Equation (13). The focused image can be obtained with the inverse Fourier
transforms of the guided wave fields at different radii.
The main steps of the proposed algorithm are described below:
Step 1: A data matrix is constructed with scanning signals collected by the transducer at the radius
R1 as column vectors.
Step 2: The two-dimensional Fourier transform of the data matrix is performed to obtain
two-dimensional spectrum P( R1 , k α , ω ).
Step 3: The length ΔR and number m of the radius step are determined; The imaging range of the
radius is between R1 and R1 − mΔR. The counter q is 1.
Step 4: Two-dimensional spectrum P( R1 − qΔR, k α , ω ) at the radius R1 − qΔR is calculated with
Equation (13). Row vector Rv ( R1 − qΔR) is constructed with the summation of each column
of the two-dimensional spectrum.
Step 5: The inverse Fourier transform of Row vector Rv ( R1 − qΔR) is performed. The result is the
qth row of the image matrix. The counter q is increased by one.
Step 6: Steps 4 and 5 are executed repeatedly until q is larger than m. The image matrix is outputted
at last.
The attainable angular resolution can be calculated with the similar derivation as in [41]. If the
divergence angle θd is used to determinetheangular coverage
of the transducer, the thresholds of the
angular wave number k α are − 4π R1 sin θ2d and 4π θd
λ R1 sin 2 , thus, the bandwidth of the angular
λ
θd
wave number Δk α is 2π
λ R1 sin 2 . Attainable angular resolution θ ar can be calculated as follows:
2π λ
θ ar = = (14)
Δk α R1 sin θ2d
194
Appl. Sci. 2018, 8, 1315
For the SH0 mode at a certain frequency, the wave length is constant, and the attainable
angular resolutions are affected by the divergence angles, which are determined by the structures of
the transducers.
The proposed method is applicable for the circumferential scanning of isotropic plates with
non-dispersive modes. For dispersive modes like the Lamb waves and higher-order SH waves,
dispersion compensation can be achieved with the relationships between the group velocities ĉ g and
the angular frequencies ω from the group velocity curves in the calculations of two-dimensional
spectrums using Equation (13). For anisotropic plates, the wave equation in Equation (8) is not suitable,
since the wave propagations along different directions are not the same. The proposed method may be
not applicable.
The detection ability of the circumferential scanning is determined by multiple reasons such as
the excitation frequency and the characteristics of the defects (shapes and sizes). With the increments
of the excitation frequencies, the detection sensitivities of the waves increase while the propagation
distances decrease. The shapes of the defects affect the mode conversions and propagations of the
scattering waves. Details can be found in [44].
3. Numerical Simulation and Verification
3.1. Simulation Setup

For simplification, only Steps 2 and 3 in Section 2.3 are carried out in the simulation. As in
Figure 7, the side length of the iron plate is 1.5 m; the thickness and material properties are the same as
in Section 2.1. The original point O2 is set in the vertex of the lower right corner; the defect is 0.825 m
from O2 with a size of 1 × 10 mm, and the angular position α1 is 10◦ . The defect is motivated with a
tone-burst signal, which is acquired by filtering a sinusoidal signal of multiple cycles with a Hanning
window, as in Figure 8. The main frequency is 128 kHz. With the angular step θs of 0.5◦ , 91 receiving
points are set at the radius 1.3 m, which are marked from RP1 to RP91. The MST is always towards
the original point during the signal being received from RP1 to RP91.The sound field of the MST for
receiving is the same as that for generation, and so the signals collected at these receiving points are
modulated by the sound field of the MST to construct the B-scan image.
Ʌ‫ݏ‬
ʹ
Ƚͳ Ƚʹ
ͳ
ൌ ͳǤͷ; ͳ ൌ ͳǤ͵; ʹ ൌ ͲǤͺʹͷ
Ƚͳ ൌ ͳͲιǢȽʹ ൌ ͶͷιǢ Ⱦ ൌ ʹʹǤͷιǢߠ‫ ݏ‬ൌ ͲǤͷι
ܱʹ
Figure 7. The diagram of the numerical model.
195
Appl. Sci. 2018, 8, 1315
(a)
(b)
(c)
(d)
Figure 8. The excitation signals with different cycles and their amplitude spectrums: (a) three cycles;
(b) five cycles; (c) seven cycles; (d) nine cycles. The half-decline frequency bandwidth of the signals
is marked.
196
Appl. Sci. 2018, 8, 1315
3.2. Method Comparison and Factor Analysis

The results of the proposed synthetic aperture focusing technology method for circumferential
scanning (CSAFT) are compared with the B-scan image and the results by the DAS method [29].
The sound field of the MST with the length 2L = 70 mm is used to modulate the signals from the
simulation. The time signals are turned into distance signals with the group velocity ĉ g by R = t × ĉ g ;
all the distance signals are stacked to become the B-scan image in the order of the angle positions
of the receiving points. By the DAS method, the image region is first divided into pixels; then, the
flight time of each pixel is calculated with the group velocity and the distance between the pixel and
the transducer in the scanning. The amplitude of each pixel is the superposition of the amplitudes
extracted from the time signals of the transducer at different positions according to the flight times.
It should be noticed that, since only the procedure of receiving is simulated, the group velocity ĉ g will
be the same as the group velocity c g = 3208 m/s. The results in Figures 9–11 are all calculated with
the excitation signal of three cycles.
(a) (b) (c)
Figure 9. The results for the angular step 0.5◦ by different methods: (a) B-scan, (b) delay and sum
(DAS), (c) circumferential synthetic aperture focusing technique (CSAFT). The excitation signal is a
tone-burst signal with three cycles.
(a) (b)
Figure 10. Comparisons of three methods with different factors: (a) the angular bandwidth θw by the
transducer with different lengths; (b) Calculation times with different numbers of signals. The excitation
signal is a tone-burst signal with three cycles.
The results for the angular step 0.5◦ are shown in Figure 9. In the B-scan image of Figure 9a,
extra waves existing in front of the defect waves are Lamb modes generated by the defect, which
travel faster than the SH0 mode. The reflection waves from the boundaries appear after the defect
waves. As in Figure 6, the divergence angle of the transducer cannot be zero, and so the defect can
also be detected even if it is not directly in front of the transducer. That is the reason for the tailings of
the defect waves in the B-scan image, which significantly lower the angular resolution. The image is
197
Appl. Sci. 2018, 8, 1315
improved by the DAS method, as in Figure 9b. However, over adjustments exist as a result, leading to
some residual trailing. The result of the CSAFT method is shown in Figure 9c. The best angular
resolution of the defect spot with the least trailing can be obtained compared with those by other
methods. The defect locations (0.76~0.77 m) in Figure 9 are calculated with the peaks of the defect
spots, which results in the localization error of the distances (6.7%) for the width of the defect waves
in the time signals. The denoted angular positions (9.5~11.5◦ ) have little error compared with the
actual value.
The effect of the length of the transducer is investigated in the simulation. The angular bandwidth
of the defect spot θw , defined as the largest angle range of the region in which the amplitudes are
no less than half of the peak, is used to denote the size of the defect spot in the angular direction.
The divergence angle θd of the transducer decreases with the increment of the length of the transducer
length 2L as in Figure 6. The angular bandwidth of the defect spot in the B-scan image has the
same tendency of decrement as the divergence angle of the transducer, as in Figure 10a, which is
12~29.5◦ for a length of the transducer between 2 and 150 mm. The DAS method provides a significant
improvement of the angular bandwidth (3.4~3.8◦ ), which is basically not affected by the variations of
the transducer lengths. The angular bandwidths by the CSAFT method (1.5~2.5◦ ) are slightly superior
to those by the DAS method. The CSAFT and DAS method are also compared for different angular
steps, which are integral multiples of the minimum angular step 0.5◦ . The numbers of the receiving
signals for different angular steps are listed in Table 2.
The effect of the excitation signals with different cycles is also studied and the results by the
CSAFT method marked with the defect positions and angular bandwidths are shown in Figure 12.
It can be observed that the widths of the defects along the radius and the angular bandwidths increase
with the increment of the cycles. The defect localization accuracy decreases as well. The increment
of the signal cycle is conducive to raising the energy of the incident waves for long testing ranges
and narrow down the frequency bandwidth to reduce the influence of other frequencies. However,
it extends the lengths of the excitation signals in the time domain, which determine the sizes of the
focused defect spots. Both resolutions along the angle and radius are decreased. The cycles of the
excitation signals should be selected in the consideration of both resolutions and testing ranges in
field tests.
The calculation time of the CSAFT and DAS methods are tested with the same number of the
signals in the same computer. As in Figure 10b, the calculation time of the CSAFT method is much less
than that of the DAS method, which increases linearly with the numbers of the signals. The advantage
of the proposed method in calculation efficiency can be more significant when dealing with a larger
amount of data. The images by three methods for different angular steps are presented in Figures 9
and 11. Since the angular sampling rate decreases with the increment of the angular step, the defect
spots are coarser in the results of the Bscan and CSAFT method. In the results by the DAS method,
over-adjustments get more severe with larger trailing in the image, which may become distractions in
defect recognition. The proposed CSAFT method provides the best performance for each angular step.
Table 2. The numbers of the receiving signals for different angular steps.
Name Value
Angular step (◦ ) 0.5 1 1.5 2
Number of the signals 91 46 31 23
198
Appl. Sci. 2018, 8, 1315
(a)
(b)
(c)
Figure 11. Comparisons of the results by the B-scan (Left), DAS (Middle), and CSAFT (Right) methods
for different angular steps: (a) θs = 1◦ ; (b) θs = 1.5◦ ; (c) θs = 2.0◦ . The excitation signal is a tone-burst
signal with three cycles.
ߠ‫ ݓ‬ൌ ʹǤͷι ߠ‫ ݓ‬ൌ Ͷι ߠ‫ ݓ‬ൌ ͷǤͷι
(a) (b) (c)
Figure 12. Comparisons of the results by the CSAFT method for excitation signals with different cycles:
(a) five cycles; (b) seven cycles; (c) nine cycles. θw is the angular bandwidth of the defect spot.
4. Experimental Investigation

Experiments were carried out to verify the proposed method. As shown in Figure 13,
the equipment named the magnetostrictive guided wave detector (MSGW30, Zheda Jingyi Tech, Ltd.,
Hangzhou, China) was used in the experiments. The equipment has the ability to generate and receive
guided waves with the modules including signal generation, power amplification, pre-amplification
199
Appl. Sci. 2018, 8, 1315
and A/D conversion under the control of the computer. The coils in the magnetostrictive sandwich
transducer (MST) were processed as a flexible printed circuit (FPC), and the material of the
magnetostrictive strip is Fe-Co alloy. The circumferential scanning is shown in Figure 14. The iron plate
is the same size as that in the simulation. An iron mass scatterer with the additional cross-sectional
area of 100 mm2 was coupled on the iron plate at the radius 0.825 m from the original point O2 and 12◦
from the start position of the scanning as the artificial defect using the epoxy coupling agent. The MST
moved around the original point O2 at a radius of 1.3 m with the angular step 1.5◦ to scan the fan area
with an angle of 45◦ . The MST was used both as an actuator and sensor. The excitation signal used in
the experiment is the same as that in the simulation, which is shown in Figure 8a. The sampling ratio
is 2 MHz. Thirty-one signals were collected in the scanning. It should be mentioned that the coupling
agent used for the coupling between the transducer and the plate is the shear wave coupler (UGW30,
Zheda Jingyi Tech, Ltd., Hangzhou, China).
Figure 13. The magnetostrictive guided wave inspection system for circumferential scanning in
the plate.
Ʌ‫ݏ‬
ʹ
Ƚͳ Ƚʹ
ͳ
ൌ ͳǤͷ; ͳ ൌ ͳǤ͵; ʹ ൌ ͲǤͺʹͷ
Ƚͳ ൌ ͳʹιǢȽʹ ൌ ͶͷιǢ Ⱦ ൌ ʹʹǤͷιǢߠ‫ ݏ‬ൌ ͳǤͷι
ܱʹ
Figure 14. Circumferential scanning in the iron plate using guided waves.
200
Appl. Sci. 2018, 8, 1315
4.2. Results and Discussion

The temperature variations should always be considered in guided wave inspections. The wave
velocities are affected by the mechanical properties of the plates, which are changed by the temperature.
Wave velocities of collected signals may be different if the scanning is carried out under large
temperature variations, affecting the quality of the scanning image. Therefore, wave velocity
compensation should be considered. In field tests, the scanning is usually finished within minutes and
the temperature does not change rapidly in a short time. However, the situation can be different when
the scanning is carried out at a different time (in the day or at night). Thus, wave velocity estimation
is necessary before scanning. In this study, the guided waves were generated directly towards one
boundary of the plate. The flight time of the reflection waves from the boundary and the distance
between the MST and the boundary were used to calculate the group velocity, which is 2714 m/s.
The reason for the error (15.4%) of the velocity may be the same as described in Section 3.2, since the
peak of the reflection wave was used to determine the flight time.
The results of the three methods are shown in Figure 15. The amplitudes of the images are all
normalized between 0 and 1. The estimated peak value of the environmental noise is less than 0.12.
The defect spot of the B-scan method in Figure 15a has large trailing, as expected, which may lower
the angular resolution. By the DAS method, the trailing around the defect spot is eliminated, as in
Figure 15b; however, some trailing occurs in other regions of the image, which have similar amplitudes
to the defect spot, becoming a severe distraction in defect recognition. In Figure 15c, the defect spot is
improved in the angular direction despite some residual trailing. It can be observed that the defect
spot by the DAS method is less obvious than that by the B-scan method, whereas the defect spot by
the CSAFT method is much more recognizable. Quantitative comparisons of the defect spot by three
methods are listed in Table 3. The position of the defect is denoted by the peak of the spot, and the
angular bandwidth is defined as in Section 3.2. The reflection waves from the boundaries are taken
as references to show the significance of the defect spots by the three methods. The peak value ratio
between the defect and the boundaries by the proposed method is 580% larger than that by the B-scan
method, whereas the result with the DAS method does not give any improvement. Apart from the
largest significance of the defect, the results by the CSAFT method have the least error in the angular
localization and smallest angular bandwidth of the defect compared with the other two methods.
(a) (b) (c)
Figure 15. Experimental images by the three methods: (a) B-scan; (b) DAS; (c) CSAFT.
Table 3. Defect estimation by the three methods.
Name Actual Values B-Scan DAS CSAFT

Angle (◦ ) 12 13.5 12.5 12
Position
Distance (m) 0.825 0.823 0.815 0.821
Angular bandwidth (◦ ) NAN 1 4.5 3.8 2.0
Peak value ratio 1
NAN 0.25 0.14 1.71
(defect/boundaries)
1 NAN means no value for this option.
201
Appl. Sci. 2018, 8, 1315
5. Conclusions
A synthetic aperture focusing technique method is proposed for circumferential scanning of
plates using guided waves. The SH0 mode is selected for the inspection, which can be effectively
generated by the magnetostrictive sandwich transducer (MST). The proposed imaging method is based
on the exploding reflector model (ERM) by solving the wave equation in a polar coordinate system
and deriving the relationships between guided wave fields at different radii in the frequency domain.
Numerical and experimental validation was successfully performed for the circumferential scanning
of the defects in an iron plate. The main findings are summarized as follows:
1. The SH0 mode has the same detection sensitivity for the cross section of the plate with a uniform
wave structure along its thickness; it is nondispersive in frequency and has little energy leakage
for the inspection of plates contacted with liquids, which can be generated purely at the frequency
below the cutoff frequency of the SH1 mode. The SH0 mode is ideal for the inspection of tank
floors in service.
2. The MST used for SH0 wave generation can control the wave propagation direction. The main
structure of the MST is designed for the SH0 mode at a certain frequency, and the divergence
angle can be adjusted with the length of the transducer.
3. The angular bandwidth of the defect spot by the proposed method (1.5~2.5◦ ) in the numerical
simulations are found to be better than those by the traditional delay and sum (DAS) method
(3.4~3.8◦ ) for transducers with different lengths (2~150 mm), both of which have a significant
improvement compared with those by the B-scan image (12~29.5◦ ). The proposed method
has better calculation efficiency than the DAS method, particularly for a large amount of data.
No additional trailing occurs in the image of the proposed method with the increment of the
angular step. In circumferential scanning, there should not be too many cycles of the excitation
signal when the testing range is sufficient.
4. By the proposed method, the peak amplitude of the defect spot can be increased, providing a
significant improvement for the defect spot compared with those by the B-scan and DAS methods.
The CSAFT method is expected to be effective for the non-destructive scanning of tank floors in
service to minimize the risks and costs of the maintenance operations. For future study, actual oil tank
floors installed with accessories will be examined. Defects of different kinds, sizes, and numbers will
be inflicted on the tank floor, which will be tested when the oil tank is empty and filled with different
kinds of liquids in the field.
Author Contributions: J.W. and S.W. conceived and designed the models and methods presented; Z.T., F.L., and
K.Y. helped with the conception and experiments; J.W. performed the simulations, experiments, and data analysis;
J.W. and Z.T. wrote this paper.
Funding: The work was supported by the National Natural Science Foundation of China (Grant Nos. 61271084,
51275454, and U1709216) and the Technique Plans of Zhejiang Province (2017C01042).
References
1. Tetteh-Wayoe, D. Shell Corrosion Allowance for Aboveground Storage Tanks. In Proceedings of the
2008 7th ASME International Pipeline Conference, Calgary, AB, Canada, 29 September–3 October 2008;
American Society of Mechanical Engineers: New York, NY, USA, 2009.
2. Yang, T.; Zhang, X.; Huang, Z.S.; Yang, F.; Zhang, T.; Wu, J.W. Vertical Tank Bottom Line Integrated Acoustic
Defect Detection Technology Research and Application. Pipeline Tech. Equip. 2016, 4, 21–23. (In Chinese)
[CrossRef]
3. Liu, Z.; Kang, Y.; Wu, X.; Yang, S. Study on local magnetization of magnetic flux leakage testing for storage
tank floors. Insight 2003, 45, 328–331. [CrossRef]
4. Ramírez, A.R.; Mason, J.S.; Pearson, N. Experimental study to differentiate between top and bottom defects
for MFL tank floor inspections. NDT E Int. 2009, 42, 16–21. [CrossRef]
202
Appl. Sci. 2018, 8, 1315
5. Usarek, Z.; Warnke, K. Inspection of Gas Pipelines Using Magnetic Flux Leakage Technology. Adv. Mater. Sci.
2017, 17, 37–45. [CrossRef]
6. Murayama, R.; Makiyama, S.; Kodama, M.; Taniguchi, Y. Development of an ultrasonic inspection robot
using an electromagnetic acoustic transducer for a Lamb wave and an SH-plate wave. Ultrasonics 2004, 42,
825–829. [CrossRef] [PubMed]
7. Kwon, J.R.; Lyu, G.J.; Lee, T.H.; Kim, J.Y. Acoustic emission testing of repaired storage tank. Int. J. Press.
Vessel. Pip. 2001, 78, 373–378. [CrossRef]
8. Riahi, M.; Shamekh, H.; Khosrowzadeh, B. Differentiation of leakage and corrosion signals in acoustic
emission testing of aboveground storage tank floors with artificial neural networks. Russ. J. Nondestruct. Test.
2008, 44, 436–441. [CrossRef]
9. Paulauskiene, T.; Zabukas, V.; Vaitiekūnas, P. Investigation of volatile organic compound (VOC) emission in
oil terminal storage tank parks. J. Environ. Eng. Landsc. Manag. 2009, 17, 81–88. [CrossRef]
10. Zhang, X.W.; Tang, Z.F.; Lv, F.Z.; Pan, X.H. Helical comb magnetostrictive patch transducers for inspecting
spiral welded pipes using flexural guided waves. Ultrasonics 2017, 74, 1–10. [CrossRef] [PubMed]
11. Zhang, X.W.; Tang, Z.F.; Lv, F.Z.; Yang, K.J. Scattering of torsional flexural guided waves from circular holes
and crack-like defects in hollow cylinders. NDT E Int. 2017, 89, 56–66. [CrossRef]
12. Sharma, S.; Mukherjee, A. Ultrasonic guided waves for monitoring corrosion in submerged plates.
Struct. Control Health Monit. 2015, 22, 19–35. [CrossRef]
13. Ostachowicz, W.; Kudela, P.; Malinowski, P.; Wandowski, T. Damage localisation in plate-like structures
based on PZT sensors. Mech. Syst. Signal. Proc. 2009, 23, 1805–1829. [CrossRef]
14. Wilcox, P.D. Omni-directional guided wave transducer arrays for the rapid inspection of large areas of plate
structures. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2003, 50, 699–709. [CrossRef] [PubMed]
15. Evans, M.; Lucas, A.; Ingram, I. The inspection of level crossing rails using guided waves. Constr. Build. Mater.
2018, 179, 614–618. [CrossRef]
16. Mazeika, L.; Kazys, R.; Raisutis, R.; Demcenko, A.; Sliteris, R.; Cantore, C. Non-Destructive Testing of
Fuel Tanks Using Long-Range Ultrasonic. In Proceedings of the 4th International Conference on Emerging
Technologies in Non-Destructive Testing, Stuttgart, Germany, 2–4 April 2007; Busse, G., VanHemelrijck, D.,
Solodov, I., Anastasopoulos, A., Eds.; Taylor & Francis Ltd.: London, UK, 2011.
17. Hay, T.R.; Royer, R.L.; Gao, H.D.; Zhao, X.; Rose, J.L. A comparison of embedded sensor Lamb wave
ultrasonic tomography approaches for material loss detection. Smart Mater. Struct. 2006, 15, 946. [CrossRef]
18. Giridhara, G.; Rathod, V.T.; Naik, S.; Mahapatra, D.R.; Gopalakrishnan, S. Rapid localization of damage using
a circular sensor array and Lamb wave based triangulation. Mech. Syst. Signal. Proc. 2010, 24, 2929–2946.
[CrossRef]
19. Rathod, V.T.; Mahapatra, D.R. Ultrasonic Lamb wave based monitoring of corrosion type of damage in plate
using a circular array of piezoelectric transducers. NDT E Int. 2011, 44, 628–636. [CrossRef]
20. Chakraborty, N.; Rathod, V.T.; Mahapatra, D.R.; Gopalakrishnan, S. Guided wave based detection of damage
in honeycomb core sandwich structures. NDT E Int. 2012, 49, 27–33. [CrossRef]
21. Ravi, N.B.; Rathod, V.T.; Chakraborty, N.; Mahapatra, D.R.; Sridaran, R.; Boller, C. Modeling Ultrasonic NDE
and Guided Wave based Structural Health Monitoring. In Proceedings of the Conference on Structural Health
Monitoring and Inspection of Advanced Materials, Aerospace, and Civil Infrastructure, San Diego, CA, USA,
9–12 March 2015; Shull, P.J., Ed.; SPIE—The Internaitional Society for Optical Engineering: Bellingham, WA,
USA, 2015.
22. Zhao, X.; Royer, R.L.; Owens, S.E.; Rose, J.L. Ultrasonic lamb wave tomography in structural health
monitoring. Smart Mater. Struct. 2011, 20, 105002. [CrossRef]
23. Zhao, X.; Rose, J.L. Ultrasonic tomography for density gradient determination and defect analysis. J. Acoust.
Soc. Am. 2005, 117, 2546. [CrossRef]
24. Zhao, X.; Rose, J.L.; Gao, F.D. Determination of density distribution in ferrous powder compacts using
ultrasonic tomography. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2006, 53, 360–369. [CrossRef] [PubMed]
25. Zhao, X.L.; Gao, H.D.; Zhang, G.F.; Ayhan, B.; Yan, F.; Kwan, C.; Rose, J.L. Active health monitoring of an
aircraft wing with embedded piezoelectric sensor/actuator network: I. Defect detection, localization and
growth monitoring. Smart Mater. Struct. 2007, 16, 1208–1217. [CrossRef]
26. Rao, J.; Ratassepp, M.; Fan, Z. Guided wave tomography based on full waveform inversion. IEEE Trans.
203
Appl. Sci. 2018, 8, 1315
27. Rao, J.; Ratassepp, M.; Fan, Z. Limited-view ultrasonic guided wave tomography using an adaptive
regularization method. J. Appl. Phys. 2016, 120, 113–127. [CrossRef]
28. Puchot, A.R.; Cobb, A.C.; Duffer, C.E.; Light, G.M. Inspection Technique for above Ground Storage Tank
Floors using MsS Technology. In Proceedings of the 10th International Conference on Barkhausen and
Micro-Magnetics, Baltimore, ML, USA, 21–26 July 2013; Chimenti, D.E., Bond, L.J., Thompson, D.O., Eds.;
American Institute Physics: New York, NY, USA, 2014.
29. Wang, C.H.; Rose, J.T.; Chang, F.K. A synthetic time-reversal imaging method for structural health monitoring.
Smart Mater. Struct. 2004, 13, 415–423. [CrossRef]
30. Pulthasthan, S.; Pota, H.R. Detection, localization and characterization of damage in plates with an, in situ
array of spatially distributed ultrasonic sensors. Smart Mater. Struct. 2008, 17, 035035. [CrossRef]
31. Sicard, R.; Goyette, J.; Zellouf, D. A saft algorithm for lamb wave imaging of isotropic plate-like structures.
Ultrasonics 2002, 39, 487–494. [CrossRef]
32. Sicard, R.; Chahbaz, A.; Goyette, J. Guided lamb waves and l-saft processing technique for enhanced
detection and imaging of corrosion defects in plates with small depth-to-wavelength ratio. IEEE Trans.
33. Davies, J.; Cawley, P. The application of synthetic focusing for imaging crack-like defects in pipelines using
guided waves. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2009, 56, 759–771. [CrossRef] [PubMed]
34. Aristegui, C.; Lowe, M.J.S.; Cawley, P. Guided waves in fluid-filled pipes surrounded by different fluids.
35. Fan, Z.; Lowe, M.J.S.; Castaings, M.; Bacon, C. Torsional waves propagation along a waveguide of arbitrary
cross section immersed in a perfect fluid. J. Acoust. Soc. Am. 2008, 124, 2002–2010. [CrossRef] [PubMed]
36. Kwun, H.; Kim, S.Y.; Light, G.M. Long-range guided wave inspection of structures using the magnetostrictive
sensor. J. Korean Soc. NDT 2001, 21, 383–390.
37. Rose, J.L. Ultrasonic Guided Waves in Solid Media, 1st ed.; Cambridge University Press: Cambridge, UK, 2014;
pp. 272–275, ISBN 978-1-107-04895-9.
38. GB 50341-2014. Code for Design of Vertical Cylindrical Welded Steel Oil Tanks. China National Petroleum
Corporation, 2012. Available online: www.csres.com/detail/243638.html (accessed on 29 May 2014).
39. Kim, I.K.; Kim, Y.Y. Shear horizontal wave transduction in plates by magnetostrictive gratings. J. Mech.
Sci. Technol. 2007, 21, 693–698. [CrossRef]
40. Lee, J.S.; Kim, Y.Y.; Cho, S.H. Beam-focused shear-horizontal wave generation in a plate by a circular
magnetostrictive patch transducer employing a planar solenoid array. Smart Mater. Struct. 2008, 18, 015009.
[CrossRef]
41. Guerra, C.; Biondi, B. Fast 3D migration-velocity analysis by wavefield extrapolation using the prestack
exploding-reflector model. Geophysics 2011, 76, WB151–WB167. [CrossRef]
42. Wu, S.W.; Skjelvareid, M.H.; Yang, K.J.; Chen, J. Synthetic aperture imaging for multilayer cylindrical object
using an exterior rotating transducer. Rev. Sci. Instrum. 2015, 86, 083703. [CrossRef] [PubMed]
43. Haun, M.A.; Jones, D.L.; O’Brien, W.D. Efficient three-dimensional imaging from a small cylindrical aperture.
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2002, 49, 1589–1592. [CrossRef]
44. Demma, A.; Cawley, P.; Lowe, M. Scattering of the fundamental shear horizontal mode from steps and
notches in plates. J. Acoust. Soc. Am. 2003, 113, 1880–1891. [CrossRef] [PubMed]
204
applied
sciences
Article
Signal Strength Enhancement of Magnetostrictive
Patch Transducers for Guided Wave Inspection by
Magnetic Circuit Optimization
Jianjun Wu 1 , Zhifeng Tang 2, *, Keji Yang 1 and Fuzai Lv 1
Zhejiang University, 38 Zheda Road, Hangzhou 310027, China; 11425023@zju.edu.cn (J.W.);
yangkj@zju.edu.cn (K.Y.); lfzlfz@zju.edu.cn (F.L.)
2 Institute of Advanced Digital Technologies and Instrumentation, School of Biomedical Engineering and
Received: 16 March 2019; Accepted: 4 April 2019; Published: 9 April 2019
Featured Application: Signal strength enhancement of the magnetostrictive patch transducer,

which can be used as an element of a phase array or scanner in pipe and plate inspection.
Abstract: Magnetostrictive patch transducers (MPT) with planar coils are ideal candidates for shear
mode generation and detection in pipe and plate inspection with the advantages of flexibility,
lightness and good directivity. However, the low energy conversion efficiency limits the application
of the MPT in long distance inspection. In this article, a method for the enhancement of the MPT was
proposed by dynamic magnetic field optimization using a soft magnetic patch (SMP). The SMP can
reduce the magnetic resistance of the magnetic circuit, which increases the dynamic magnetic field
intensity in the magnetostrictive patch during wave generation and restricts the induced dynamic
magnetic field within the area around the coils for sensing during wave detection. Numerical
simulations carried out at different frequencies verified the improvement of the dynamic magnetic
fields by the SMP and influence of different affecting factors. The experimental validations of
the signal enhancement in wave generation and detection were performed in an aluminum plate.
The amplitude magnification could reach 12.7 dB when the MPTs were covered by the SMPs. Based
on the numerical and experimental results, the SMP with a large relative permeability and thickness
and close fitting between the SMP and coils were recommended when other application conditions
were met.
Keywords: magnetostrictive patch transducer; shear mode; soft magnetic patch; dynamic magnetic
field optimization; signal strength enhancement
1. Introduction
Guided waves, which are ultrasonic waves restricted by the boundaries of the solid waveguide
during propagation [1], have been applied in the inspection of many industrial structures, such as
pipes [2,3], rods [4], plates [5–7] and rails [8–10] with the advantages of long distance, full coverage
and low cost [11]. The conversions between the electrical energy and mechanical energy are mainly
involved in guide wave inspection, which are accomplished by the transducers as the key parts
in the inspection system. Transducers with various characteristics, including vibration directions,
frequency bandwidths and sizes, are generally customized to satisfy the demands under different
conditions. Based on the physical effects, the guided wave transducers can be categorized into
different types, including electromagnetic acoustic transducers (EMAT), laser transducers, piezoelectric

Appl. Sci. 2019, 9, 1477
transducers (PZT) and magnetostrictive transducers. The EMATs are able to work without contacting
the waveguide using the Lorentz force [11,12] although their applications are limited due to the
low conversion efficiency, inapplicability for non-metal waveguides and signal instability caused by
the coil vibration. The laser can be used to generate guided waves in a contactless manner based
on the thermoelastic effect. However, the high power laser may burn the spot in the waveguide
and the generated waves normally have very high frequencies (several MHz). With the high price,
the inspection system with laser transducers still remain in laboratories without being widely used
so far [13,14]. The magnetostrictive transducers and PZTs are widely used at a low cost and with
satisfactory performance [15]. The magnetostrictive transducers can be made into small patches,
which are flexible and light for different conditions [16]. Compared with the PZTs, they have the
advantages of good durability, low requirement in coupling, low cost and superiority in shear vibration
generation. For the inspection of ferromagnetic materials, the magnetostrictive transducers can also be
used without contact [17].
The dimensions of the ferromagnetic materials change during the process of magnetization
and their magnetic state varies when external stress is applied to them. These effects are from
the properties of ferromagnetic materials, which are called the magnetostrictive effect (Joule effect)
and inverse magnetostrictive effect (Villari effect), respectively. These effects can be explained with
magnetic domain theory [18]. The magnetostrictive effect works in the generation of guided waves
while the inverse effect accounts for the measurement [19]. Many studies have been carried out for
the improvement of energy conversion, directivity and mode selection since the first application of
the magnetostriction in guided wave inspection. In early studies, the magnetostriction (λs ) of the
ferromagnetic waveguide was used to generate guided waves [20]. This method still works in cable
inspections [21,22]. However, the magnetostriction of these ferromagnetic waveguides, which are
commonly made of iron (λs = 14 ppm), are very small and thus, the patches and strips made of
materials with a large magnetostriction, such as Ni (λs = −50 ppm), Co (λs = −93 ppm) and Hiperco
(λs = 60~100 ppm), are coupled with the waveguide for wave generation and detection [23–25],
which can also be used for non-ferromagnetic waveguides [26]. There has been research carried out
on the application of giant magnetostrictive materials, such as Galfenol (λs = 200~400 ppm) and
Terfenol-D (λs = 1500~2000 ppm) in guided wave inspection [27,28]. These materials have poor
machinability and need large driving magnetic fields for an ideal performance. The undesirable
temperature adaptability and high cost also restricts their applications in field tests. There are two
magnetic fields named the static bias magnetic field and dynamic alternating field, which should be
applied into the magnetostrictive patches during the wave generation. The bias magnetic field for the
elimination of the frequency doubling effect can be provided by permanent magnets [29]. For pipe
inspection, the bias magnetic field may be more utilized when the magnetostrictive patches are bonded
with an angle of 45◦ to the longitudinal axis of the pipes, which can be improved further with the
patches in the Z-shape or with V-shaped yokes [30,31]. For magnetostrictive materials with certain
coercive forces (e.g., Hiperco), the bias magnetic field should be more uniform, which can be achieved
by pre-magnetization [32]. A solenoid coil can also provide a strong bias magnetic field, but a large
direct current is needed and the magnetostrictive strip needs to be manufactured with grooves for
solenoid coils [33]. The dynamic magnetic field is normally provided by the coils. Figure 1 shows
the common coils used in transducers with directivity. The figure of eight coil has a large energy
conversion efficiency with multiple turns as well as the coil with a core [34]. However, a high system
power is needed to drive these coils with large resistances and the directivities are not satisfactory.
The transducers with these coils are too large for narrow spaces and cannot be deformed to match
curving surfaces. Planar coils in the shape of meanders or racetracks have good directivities. They
can be manufactured as flexible printed circuits (FPC), which are small, light and flexible. As a result,
they can be used as scanners in pipe and plate inspections or elements in phased arrays. The distance
between the adjacent wires in planar coils depends on the generation of the wavelength from the mode,
resulting in the sparse arrangement of the wires. Furthermore, these coils work in open magnetic
206
Appl. Sci. 2019, 9, 1477
circuits with dynamic magnetic fields dispersed in the air [19]. These characteristics result in the
relative low energy conversion efficiency of the MPT with planar coils, limiting their applications in
long distance inspection.
Figure 1. Coils commonly used in the magnetostrictive transducers.
In this paper, a method for the enhancement of the magnetostrictive patch transducer (MPT)
with planar coils was proposed by the dynamic magnetic field optimization using a soft magnetic
patch (SMP). In Section 2, the relationship between the magnetic fields and mechanical properties of
the magnetostrictive materials is illustrated and the theoretical explanations of the improvement for
the dynamic magnetic field are given based on the magnetic Ohm’s law. As described in Section 3,
a 2D model of the transducer was built to verify the dynamic magnetic field improvement and to
study the effect of different affecting factors, including the thickness, magnetic conductivity, lift-off
distance of the SMP with the excitation currents and induced magnetic fields at different frequencies.
Experimental investigations were carried out in an aluminum plate as described in Section 4 and the
conclusions follow in Section 5.
2. Theory
2.1. Magnetostrictive Equation

In ferromagnetic materials, adjacent atoms construct magnetic domains spontaneously because
of the forces caused by electron spins [18]. As shown in Figure 2, without external magnetic fields,
the magnetic domains distribute randomly, which results in the material having a total magnetic
intensity of zero. When the external magnetic field is nonzero, the magnetic domains will move and
spin to be parallel to it, changing the length of the object along the same direction. The magnetic
domain theory gives a qualitative explanation of the magnetostrictive effect. The constitutive equations
of magnetic and elastic properties in ferromagnetic materials are provided as follows [35]:
S = sσ + dH (1)
σ = cS − eH (2)
B = d σ + μH
T
(3)
where the superscript T denotes the transposition; S and σ are the strain and stress, respectively;
H is the magnetic field intensity; B is the magnetic flux density; s and c are the elastic compliance
matrix and stiffness matrix, respectively; d is the piezomagnetic matrix; e is the inverse matrix; and μ
is the permeability. Equations (1) and (2) account for the magnetostrictive effect while Equation (3)
contributes to the inverse magnetostrictive effect. Figure 3 shows the diagram of a MPT with planar
coils. The static bias magnetic field in the magnetostrictive patch was provided by pre-magnetization,
making the structure of the MPT thin and flexible without permanent magnets. The dynamic magnetic
field was supplied by the coils sealed in the FPC, which sensed induced dynamic magnetic fields
during the wave detection. In this paper, the model of the MPT was considered for the generation and
detection of wave modes with shear vibrations in plates (shear horizontal waves) and pipes (torsional
waves), which had the advantage of non-dispersion at low frequencies for inspection. The models for
the longitudinal wave MPTs can be found in [36]. In shear mode generation, the dynamic magnetic
207
Appl. Sci. 2019, 9, 1477
field is perpendicular to the static magnetic field, of which the intensity is much smaller than that of
the static magnetic field. The total strain is caused by the field superposition:
St = sσt + d(Hd + Hs ) (4)
where ⎡ ⎤T
3ε M
0 0 0 0 0 Hs
⎢ ⎥
d = ⎣ − ϕ2 ϕ − ϕ2 0 0 0 ⎦ (5)
3ε M
0 0 0 Hs 0 0
σt is the applied stress; Hd and Hs denote the intensities of the dynamic and static magnetic fields,
respectively; ϕ is the gradient of the magnetostrictive curve of the material at Hs ; and ε M is the total
magnetostrictive strain, which mostly depends on the static magnetic field since the dynamic field is
much smaller. Considering that the wave amplitude is correlated with the dynamic shear strain Sds ,
Equation (4) can be simplified as:
3ε
Sds = M |Hd | (6)
|Hs |
It can be concluded from Equation (6) that the dynamic shear strain is proportional to the
amplitude of the dynamic magnetic field. Thus, it may be an effective way to raise the wave energy by
enhancing the dynamic magnetic field.
Figure 2. Magnetic domain movement in the magnetostrictive effect.
Figure 3. Magnetostrictive patch transducer for shear wave inspection.
2.2. Enhancement with a Soft Magnetic Patch

In wave generation, the dynamic magnetic field in the magnetostrictive patch is supplied by the
coils loaded with sinusoidal pulses. It can be described with the magnetic Ohm’s law qualitatively.
As shown in Figure 4, the coils in the middle were the magnetic source, which supplied a constant
magnetomotive force F when the input power of the equipment stayed unchanged. The magnetic flux
circulated around the coils constructing a space named the magnetic circuit. According to the magnetic
Ohm’s law, the relation between the magnetomotive force F and the magnetic flux ∅ is as follows:
208
Appl. Sci. 2019, 9, 1477
F
∅= , (i = 1, 2, 3, 4) (7)
∑4i=1 Ri
where
Li
Ri = , (i = 1, 2, 3, 4) (8)
μ i μ0 A i
Ri is the magnetic resistance of part i in the magnetic circuit as marked in Figure 4; Li and Ai
is the length and sectional area, respectively; and μi is the relative permeability. Furthermore,
μ0 = 4π × 10−7 N ·A−2 is the permeability of the vacuum. Part 1 denotes the magnetostrictive patch.
Parts 2, 3 and 4 were full of air. The magnetostrictive patch was not magnetized along the direction of
the dynamic magnetic field and thus, the magnetic intensity Hd in part 1 could be calculated as:
∅
Hd = (9)
μ1 μ0 A1
The magnetic resistance of parts 2 and 4 were the same (i.e., R2 = R4 ). Combined with
Equations (7) and (8), Equation (9) can be written as follows:
F
Hd = μ A μ1 A1
(10)
L1 + 2 μ12 A12 L2 + μ3 A3 L 3
It could be found that without changing the material property (μ1 ) and sizes (L1 and A1 ) of
the magnetostrictive patch (part 1), the magnetic intensity Hd could be improved with an increase
in the relative permeability of part 3 (μ3 ). It could be achieved by covering a soft magnetic patch
(SMP) over the coils. Hd could be improved further by increasing the thickness of the SMP (related
to A3 ) and decreasing the gap (L2 ) between the SMP and magnetostrictive patch. It should be
noted that Equation (10) qualitatively explains the improvement by the SMP in wave generation.
The magnetic flux was distributed unevenly in the magnetic circuit, which also does not hold a regular
geometric structure.
Figure 4. Diagram of dynamic magnetic fields in the MPTs (magnetostrictive patch transducers).
In wave detection, the magnetostrictive patch becomes a magnetic source because of the inverse
magnetostrictive effect. The magnetic field induced by the magnetostrictive patch was proportional
to the wave strength without additional magnetic fields. With the SMP above the coils, the induced
magnetic field may have been restricted within the area around the coils and the induced current
in the coils would be larger with an amplitude increase in the induced dynamic magnetic field.
The improvement of the SMP can affect both processes when a MPT is used as an actuator and sensor
at the same time. Furthermore, the electromagnetic interference, which are ubiquitous in the testing
field, can be blocked by the SMP, improving the signal-to-noise ratio (SNR) of the detection signals.
209
Appl. Sci. 2019, 9, 1477
3. Numerical Simulation
3.1. Simulation Setup

The dynamic magnetic field in the working region of the MPTs can be regarded as being uniform
along the direction of the wires. Thus, for simplification, a two-dimensional model was built in
COMSOL Multiphysics (COMSOL 5.3a, COMSOL Inc., Stockholm, Sweden) as shown in Figure 5.
The coupling layer between the magnetostrictive patch and aluminum plate was neglected as it was
non-conducting and non-magnetic. The coils of the two layers sealed in the FPC were made of copper.
The material of the magnetostrictive patch was Fe–Co alloy. Permalloy, silicon steel, pure iron and
Mn–Zn ferrite were the optional materials with a high relative permeability for the SMP. Considering
the eddy-current loss, which may be caused by the dynamic magnetic field, the Mn–Zn ferrite with
low conductivity was selected. In Figure 5, the real sizes of different parts are marked. The symbols
represent the dimensions in the model (l: Length; t: Thickness; g: Gap distance) and the subscripts
denote different parts (a: Aluminum plate; m: Magnetostrictive patch; c: Coil; s: SMS). lcc was the
interval length of wires in the same coil. The material properties are listed in Table 1. In this study,
the generation and detection of the zeroth shear horizontal mode (SH0 mode) was considered to
have the advantage of non-dispersion. To avoid the interference of high order modes, the excitation
frequency should be below the cutoff frequency of the SH1 mode (312 kHz) in the aluminum plate
(the thickness was 5 mm). With the consideration of the high attenuation of wave modes at high
frequencies and low detection sensitivity at low frequencies in field tests, the frequency range was
determined as 55–150 kHz. The frequency tests were carried out with increasing steps of 5 kHz
using the FPCs named MF64 and MF128, which were designed at the main frequencies of 64 kHz
and 128 kHz, respectively (MF64 for 55–90 kHz, MF128 for 95–150 kHz). Two coils were designed
in one FPC for wave direction controlling as shown in Figure 3, but only one was used to avoid
channel crosstalk in experiments. The simulations used the same setup and the current directions
are shown in Figure 6. The red solid lines in Figure 5 are reference lines with the length of lm to
observe the magnetic intensities. For wave generation, the alternating currents with the amplitude
1A were loaded into the coils and the magnetic intensity along the x-axis of Line 1, which is in the
middle of the magnetostrictive patch, was observed. In wave detection, the alternating magnetic
fields with the amplitude 1 A/m2 are applied into the magnetostrictive patch along the x-axis and the
magnetic intensity in Line 2, which was across the middle of the coils near the magnetostrictive patch,
was observed.
OP PP OF PP OFF PP

WD PP WP PP WF PP WV PP
JPD PP JFP PP JFF PP JVF PP
Figure 5. Numerical model for dynamic magnetic field simulation of magnetostrictive patch transducers.
210
Appl. Sci. 2019, 9, 1477
Table 1. Material properties in the simulation.
Material Name Air Aluminum Copper Fe–Co Alloy Mn–Zn Ferrite

Relative permeability 1 1 1 2000 2800
Conductivity (S/m) 0 3.774 × 107 5.988 × 107 2.5 × 106 0.17
Figure 6. Current directions for different types of magnetostrictive patch transducers.
3.2. Simulation Verification and Factor Analysis

The typical dynamic magnetic field distributions of MPTs with and without the SMPs at 85 kHz
in wave generation and detection are shown in Figure 7. With the SMP, the magnetic flux lines can be
restricted in the narrow area around the coils instead of dispersing in the air region. Figure 8 shows
the amplitudes of the magnetic field intensities along the x-axis from Line 1 and Line 2, respectively.
The dynamic magnetic field existing in the magnetostrictive patch are right below the coils with
currents. In wave detection, the high peaks are magnetic intensities in the coils. The dynamic
magnetic field in the magnetostrictive patch was enhanced more than three times with the SMP and
the induced magnetic fields in the coils were also stronger. The results shown in Figure 9 confirmed the
enhancement of the SMPs at other frequencies. It should be mentioned that the sharp fall in Figure 9a
may have been caused by different current loadings as shown in Figure 6.
:DYHJHQHUDWLRQ :DYHGHWHFWLRQ

:LWKRXW603

î

:LWK603

Figure 7. Magnetic field distribution in magnetostrictive patch transducers at 85 kHz. The magnetic
lines are green lines in the case of wave generation and blue lines of wave detection.

0DJQHWLFLQWHQVLW\DORQJ[D[LV$P
:LWK603 :LWK603
:LWKRXW603 :LWKRXW603

3RVLWLRQPP 3RVLWLRQPP
(a) (b)
Figure 8. Comparisons of magnetostrictive patch transducers with and without the SMPs at 85 kHz:
(a) The magnetic field intensity in Line 1 in wave generation; (b) the magnetic field in Line 2 in
wave detection.
211
Appl. Sci. 2019, 9, 1477
:LWK603
:LWKRXW603

)UHTXHQF\N+]
(a) (b)
Figure 9. Comparisons of MPTs with and without the SMPs at 55~150 kHz: (a) The magnetic field
intensity in Line 1 in wave generation; and (b) the magnetic field in Line 2 in wave detection.
For the proper selection of the SMP, several properties, including the relative permeability,
thickness (ts ) and lift-off distance (gsc ), were studied with the other parameters of the model having
fixed values. The typical results at 55 kHz, 85 kHz, 115 kHz and 145 kHz are shown in Figures 10–12.
The magnetic fields get stronger with an increase in the relative permeability of the SMP, as shown
in Figure 10a, which matches the theoretical prediction in Section 2.2. However, when the relative
permeability is larger than 2000, the influence of relative permeability changes is subtle. The lower
limit of the relative permeability for maximum magnetic field intensity in wave detection is 100.
The dynamic magnetic field attenuates exponentially during the propagation in a good conductor so it
is only distributed near the surface of the conductor (i.e., skin effect [37]). With this effect, the thickness
of the SMP can be very small (0.01~0.6 mm in the simulation). The improvement of magnetic fields
by an increase in the thickness was verified from the curves of 55 kHz and 85 kHz in Figure 11a,
which did not increase significantly when the thickness was over 0.15 mm. For other cases at higher
frequencies or in wave detection, it reaches the maximum for higher frequencies at the minimum
thickness. The distance between the SMP and coils (i.e., the lift-off distance in this paper) will affect the
improvement of the SMP if something necessarily has to be put in between (e.g., a second insulation
layer). The influence of the SMP gets smaller naturally when the SMP gets further away from the coils.
It can be explained with the increment of the magnetic resistance in the magnetic circuits over the
lift-off distance as described in Section 2.2. Figure 12 shows the results for different lift-off distances.
It could be found that during the lift-off of the SMP, the influence attenuation of the SMP continued
until a long distance (2.5 mm) in wave generation, but the influence remained constant when the SMP
was 0.5 mm away in wave detection.
(a) (b)
Figure 10. MPTs with the SMPs of different relative permeabilities: (a) The magnetic field intensity in
Line 1 in wave generation; and (b) the magnetic field intensity in Line 2 in wave detection.
212
Appl. Sci. 2019, 9, 1477
(a) (b)
Figure 11. MPTs with the SMPs of different thicknesses: (a) The magnetic field intensity in Line 1 in
wave generation; and (b) the magnetic field intensity in Line 2 in wave detection.
I N+] I N+] I N+] I N+]

/LIWRIIGLVWDQFHPP
(a) (b)
Figure 12. MPTs with the SMPs of different lift-off distances: (a) The magnetic field intensity in Line 1
in wave generation; and (b) the magnetic field intensity in Line 2 in wave detection.
4. Experimental Investigation

Experiments were carried out to verify the improvement of the SMP as shown in Figure 13.
The equipment used in the experiments (MSGW30, Zheda Jingyi Tech, Ltd., Hangzhou, China) was
designed for magnetostrictive guided wave inspection, which was integrated with modules, such
as signal generation, power amplification, pre-amplification and A/D conversion. The equipment
has two channels. Under the control of the computer, the equipment can work in the pulse-echo
mode (wave generation and detection using Channel 1) and pitch-catch mode (wave generation using
Channel 1 and detection using Channel 2). The inspection subject was an aluminum plate with a
thickness of 5 mm. For the constant and same bonding conditions of the transducers (T1 and T2),
dual adhesive tapes were used for the bonding between the magnetostrictive patches made of Fe-Co
alloy and the aluminum plate. The thicknesses of the magnetostrictive patch and dual adhesive tape
were 0.12 mm and 0.15 mm, respectively. The FPCs (MF64 and MF128) for the two main frequencies
(64 kHz and 128 kHz) were processed with two coils and only one was used in the experiments to
avoid the crosstalk. The generated guided wave propagated forward and backward without direction
controlling and the wave packs from the far end of the plate in the signals were used for evaluation.
The SMP was manufactured by hot-pressing with a mixture of the Mn–Zn ferrite power and rubber,
which had a thickness and relative permeability of 0.15 mm and 2800, respectively. The SMP, FPC and
magnetostrictive patch were all flexible and the dimensions of the SMP and magnetostrictive patch
were 50.8 × 70 mm, which was the same as the working region of the FPC. The excitation signals at
different frequencies were sinusoidal signals of four cycles filtered by the hanning windows and the
sample frequency was 2 MHz.
213
Appl. Sci. 2019, 9, 1477
Figure 13. The diagram of the experiments for transducer improvement with the SMP.
The improvement of the SMP for wave generation and detection was first verified. In this
experiment, the transducer marked as T1 in Figure 13 was used as an actuator while T2 was a sensor.
The MF64 was used for frequencies of 55–90 kHz while the MF128 was used for frequencies of
95–150 kHz. The signals were set as the base data when T1 and T2 both were not covered by the SMPs.
Furthermore, the signals received at the cases, such as only T1 covered by the SMP, only T2 covered
by the SMP and both covered by the SMPs, were compared with the base data. The magnification
was calculated using the peak values of the wave packs from the far end of the plate in the Hilbert
envelopes of the signals.
The effect of thickness changes was investigated using multiple layers of SMPs. For each frequency,
the signals were recorded when the MPT was covered by different layers of the SMPs (four layers
at most). The last experiment was focused on the lift-off distance, which was carried out using
a single layer of the SMP. The lift-off was achieved by installing multiple layers of epoxy patches
between the FPC and SMP (six layers at most). The thickness of one single epoxy patch was 0.5 mm.
These experiments were carried out with T1 used as an actuator and sensor in the pitch–catch mode.
The signals from the cases in which the transducer was not covered by the SMP were used as the base
data for amplification evaluation. It should be mentioned that in order to get rid of the void between
the layers of the transducer, a nonmetal heavy weight was set above the transducer during the signal
collection. The magnetostrictive patch was re-magnetized at the beginning of every experimental case
to keep the uniformity of the bias magnetic field and multiple signals were collected in each case to
reduce the effect of random noises.
4.2. Results and Discussions

The original signals were all processed with band-pass filters. Figure 14 shows the signals at
85 kHz when both the actuator and sensor were or were not covered by the SMPs. Comparisons
were made between the ratios of the flight times in the signals and propagation distances from the
transducers to the near end and far end. It can be confirmed that the wave packets at 2.25 × 10−4 s
are echo waves from the near end and those at 5.48 × 10−4 s were from the far end. The far end echo
waves were observed in the experiments, of which the peak values with the SMP were about two times
of those without the SMP. The magnifications of signals at different frequencies were calculated by
using the peak values of the far end echo waves in the envelope signals, which are shown in Figure 15.
The SMP provided the enhancement for the MPT used in wave generation, which were even more
than that in wave detection. The signal strength could be improved by 9.8 dB at most with both of the
transducers covered by the SMPs.
Figure 16 shows the results of the MPT covered by different layers of SMPs at different frequencies.
The enhancement was promoted by multiple layers of SMPs, which reached a maximum amplification
of 12.7 dB at 130 kHz. However, there were no significant changes when the layer number of the
SMPs was over two. The step change at the frequencies of 90 kHz and 95 kHz may have been caused
214
Appl. Sci. 2019, 9, 1477
by two different types of FPCs in the experiments. Considering these results, the SMP with the
thickness over 0.3 mm should be recommended. The typical results for different lift-off distances of
the SMP are shown in Figure 17. The decreasing trends at different frequencies match the prediction.
The magnifications decrease faster at higher frequencies. A slight enhancement of 2.8 dB still exists
when the SMP is moved 3 mm away from the FPC. However, for better results, the SMP should have a
closer fit to the FPC under allowed conditions.
Figure 14. Original signals of the cases in which T1 and T2 both were with and without the SMPs at
85 kHz. T1 was an actuator and T2 was a sensor.
$PSOLWXGHPDJQLILFDWLRQG%
Figure 15. Amplitude magnification for the MPTs in wave generation and detection with the SMPs at
different frequencies. T1 was an actuator and T2 was a sensor.

)RXUOD\HUV
7KUHHOD\HUV
7ZROD\HUV
2QHOD\HU

)UHTXHQF\N+]
Figure 16. Amplitude magnifications for the MPT with multiple layers of SMPs at different frequencies.
T1 was an actuator and sensor.
215
Appl. Sci. 2019, 9, 1477
$PSOLWXGHPDJQLILFDWLRQG%
Figure 17. Amplitude magnification for the magnetostrictive patch transducer with different lift-off
distances of the SMP at several typical frequencies. T1 was an actuator and sensor.
5. Conclusions
A method of signal strength enhancement was proposed for the magnetostrictive patch transducer
(MPT) by dynamic magnetic field optimization with a soft magnetic patch (SMP) in guided wave
inspection. The enhancement was achieved by decreasing the resistance of the magnetic circuit in
wave generation and restricting the induced dynamic magnetic field in wave detection. A numerical
investigation was successfully performed for the improvement of dynamic magnetic fields in the
magnetostrictive patch and region around the coils. Experimental validation was carried out for the
SH0 wave inspection in an aluminum plate. The main findings are summarized as follows:
1. The amplitude enhancement of the dynamic magnetic field in the magnetostrictive patch is
beneficial for SH wave generation and the concentration of the induced magnetic field by the
magnetostrictive patch in wave detection is good for the measurement by the coils.
2. The SMP can significantly decrease the resistance of the magnetic circuit for the dynamic
magnetic field in wave generation and concentrates the induced magnetic field in wave detection.
The relative permeability should be more than 2000 and the thickness should be over 0.15 mm.
The increased distance between the SMP and the coils was bad for the improvement.
3. Covered by the SMPs with the relative permeability of 2800, the MPTs provided a better
performance in both processes of wave generation and detection compared to those without the
SMPs. The improvement of the SMP in wave detection was more than that in wave generation.
The largest magnification can be 12.7 dB when the transducer is used as an actuator and sensor at
the same time.
4. For ideal results, using the SMPs with the relative permeability of 2800, a thickness over 0.3 mm
and close fitting between the SMP and coils is recommended.
The SMP is expected to improve the signal strength of the MPT for guided wave inspection
without sacrificing the flexibility and lightness of the transducer or increasing the system power.
For future studies, the SMPs made of different materials with different relative permeabilities will
be tested and different types of environmental factors, including temperature and electromagnetic
disturbing, will be investigated. The MPT with the SMP will also be examined in field tests of pipe
and tank floor scanning.
Author Contributions: J.W. conceived and designed the models and methods presented; Z.T., F.L. and K.Y. helped
with the conception and experiments; J.W. performed the simulations, experiments and data analysis; J.W. and
Z.T. wrote this paper.
Funding: The work was supported by the National Natural Science Foundation of China (Grant No. 51875511)
and National Key Research and Development Program of China (Grant No. 2018YFC0809000).
216
Appl. Sci. 2019, 9, 1477
References
1. Rose, J.L. Ultrasonic Guided Waves in Solid Media, 1st ed.; Cambridge University Press: Cambridge, UK, 2014;
p. 1. ISBN 978-1-107-04895-9.
2. Heinlein, S.; Cawley, P.; Vogt, T.K. Reflection of torsional T (0, 1) guided waves from defects in pipe bends.
NDT E Int. 2018, 93, 57–63. [CrossRef]
3. Leinov, E.; Lowe, M.J.S.; Cawley, P. Investigation of guided wave propagation and attenuation in pipe buried
in sand. J. Sound Vib. 2015, 347, 96–114. [CrossRef]
4. Hayashi, T.; Tamayama, C.; Murase, M. Wave structure analysis of guided waves in a bar with an arbitrary
cross-section. Ultrasonics 2006, 44, 17–24. [CrossRef] [PubMed]
5. Seung, H.M.; Park, C.I.; Kim, Y.Y. An omnidirectional shear-horizontal guided wave EMAT for a metallic
plate. Ultrasonics 2016, 69, 58–66. [CrossRef] [PubMed]
6. Hasanian, M.; Lissenden, C.J. Second order harmonic guided wave mutual interactions in plate: Vector
analysis, numerical simulation, and experimental results. J. Appl. Phys. 2017, 122, 084901. [CrossRef]
7. Yu, X.; Fan, Z.; Castaings, M.; Biateau, C. Feature guided wave inspection of bond line defects between a
stiffener and a composite plate. NDT E Int. 2017, 89, 44–55. [CrossRef]
8. Wu, J.; Tang, Z.; Lv, F.; Yang, K.; Yun, C.B. Ultrasonic guided wave-based switch rail monitoring using
independent component analysis. Meas. Sci. Technol. 2018, 29, 115102. [CrossRef]
9. Evans, M.; Lucas, A.; Ingram, I. The inspection of level crossing rails using guided waves. Constr. Build.
Mater 2018, 179, 614–618. [CrossRef]
10. Yao, W.; Sheng, F.; Wei, X.; Zhang, L.; Ynag, Y. Propagation characteristics of ultrasonic guided waves in
continuously welded rail. Modern Phys. Lett. B 2017, 31, 1740075. [CrossRef]
11. Nakamura, N.; Ogi, H.; Hirao, M. EMAT pipe inspection technique using higher mode torsional guided
wave T (0, 2). NDT E Int. 2017, 87, 78–84. [CrossRef]
12. Kogia, M.; Gan, T.H.; Balachandran, W.; Livadas, M.; Kappatos, V.; Szabo, I.; Mohimi, A.; Round, A. High
temperature shear horizontal electromagnetic acoustic transducer for guided wave inspection. Sensors 2016,
16, 582. [CrossRef]
13. Lee, J.R.; Chia, C.C.; Park, C.Y.; Jeong, H. Laser ultrasonic anomalous wave propagation imaging method
with adjacent wave subtraction: Algorithm. Opt. Laser Technol. 2012, 44, 1507–1515. [CrossRef]
14. Lee, J.R.; Chia, C.C.; Shin, H.J.; Park, C.Y.; Yoon, D.J. Laser ultrasonic propagation imaging method in the
frequency domain based on wavelet transformation. Opt. Lasers Eng. 2011, 49, 167–175. [CrossRef]
15. Sevillano, E.; Sun, R.; Perera, R. Damage detection based on power dissipation measured with PZT sensors
through the combination of electro-mechanical impedances and guided waves. Sensors 2016, 16, 639.
[CrossRef]
16. Vinogradov, S.; Cobb, A.; Fisher, J. New Magnetostrictive Transducer Designs for Emerging Application
Areas of NDE. Materials 2018, 11, 755. [CrossRef]
17. Kwun, H.; Bartels, K.A. Magnetostrictive sensor technology and its applications. Ultrasonics 1998, 36, 171–178.
[CrossRef]
18. Wang, W.B.; Cao, S.Y.; Huang, W.M. Magnetostrive Materials and Devices, 1st ed.; Metallurgical Industry Press:
Beijing, China, 2008; pp. 8–10. ISBN 9787502445324. (In Chinese)
19. Kim, Y.Y.; Kwon, Y.E. Review of magnetostrictive patch transducers and applications in ultrasonic
nondestructive testing of waveguides. Ultrasonics 2015, 62, 3–19. [CrossRef]
20. Tzannes, N.S. Joule and Wiedemann effects-The simultaneous generation of longitudinal and torsional stress
pulses in magnetostrictive materials. IEEE Trans. Sonics Ultrason. 1966, 13, 33–40. [CrossRef]
21. Xu, J.; Wu, X.; Cheng, C.; Ben, A. A magnetic flux leakage and magnetostrictive guided wave hybrid
transducer for detecting bridge cables. Sensors 2012, 12, 518–533. [CrossRef]
22. Xu, J.; Wu, X.; Sun, P. Detecting broken-wire flaws at multiple locations in the same wire of prestressing
strands using guided waves. Ultrasonics 2013, 53, 150–156. [CrossRef]
23. Park, C.; Kim, Y.Y. Nonferromagnetic material inserted magnetostrictive patch bonding technique for
torsional modal testing of a ferromagnetic cylinder. Rev. Sci. Instrum. 2010, 81, 035103. [CrossRef]
24. Kumar, K.S.; Murthy, V.; Balasubramaniam, K. Improvement in the signal strength of magnetostrictive
ultrasonic guided wave transducers for pipe inspection using a soft magnetic ribbon-based flux concentrator.
Insight 2012, 54, 217–220. [CrossRef]
217
Appl. Sci. 2019, 9, 1477
25. Kim, H.W.; Cho, S.H.; Kim, Y.Y. Analysis of internal wave reflection within a magnetostrictive patch
transducer for high-frequency guided torsional waves. Ultrasonics 2011, 51, 647–652. [CrossRef]
26. Cho, S.H.; Lee, J.S.; Kim, Y.Y. Guided wave transduction experiment using a circular magnetostrictive patch
and a figure-of-eight coil in nonferromagnetic plates. Appl. Phys. Lett. 2006, 88, 224101. [CrossRef]
27. Yoo, B.; Na, S.M.; Flatau, A.B.; Pines, D.J. Ultrasonic guided wave sensing performance of a magnetostrictive
transducer using Galfenol flakes-polymer composite patches. J. Appl. Phys. 2015, 117, 17A916. [CrossRef]
28. Kim, Y.; Kim, Y.Y. A novel Terfenol-D transducer for guided-wave inspection of a rotating shaft. Sens. Actuators
A Phys. 2007, 133, 447–456. [CrossRef]
29. Cho, S.H.; Kim, H.W.; Kim, Y.Y. Megahertz-range guided pure torsional wave transduction and experiments
using a magnetostrictive transducer. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2010, 57, 1225–1229. [CrossRef]
30. Park, C.I.; Cho, S.H.; Kim, Y.Y. Z-shaped magnetostrictive patch for efficient transduction of a torsional wave
mode in a cylindrical waveguide. Appl. Phys. Lett. 2006, 89, 174103. [CrossRef]
31. Park, C.I.; Kim, W.; Cho, S.H.; Kim, Y.Y. Surface-detached V-shaped yoke of obliquely bonded
magnetostrictive strips for high transduction of ultrasonic torsional waves. Appl. Phys. Lett. 2005, 87,
224105. [CrossRef]
32. Zhang, X.; Tang, Z.; Lv, F.; Pan, X. Helical comb magnetostrictive patch transducers for inspecting spiral
welded pipes using flexural guided waves. Ultrasonics 2017, 74, 1–10. [CrossRef]
33. Kim, Y.G.; Moon, H.S.; Park, K.J.; Lee, J.K. Generating and detecting torsional guided waves using
magnetostrictive sensors of crossed coils. NDT E Int. 2011, 44, 145–151. [CrossRef]
34. Lee, J.S.; Kim, Y.Y.; Cho, S.H. Beam-focused shear-horizontal wave generation in a plate by a circular
magnetostrictive patch transducer employing a planar solenoid array. Smart Mater. Struct. 2008, 18, 015009.
[CrossRef]
35. Huang, S.L.; Wang, K.; Zhao, W. Theory and Application of Electromagnetic Ultrasonic Guided Wave, 1st ed.;
Tsinghua University Press: Beijing, China, 2013; p. 74. ISBN 978-7-302-33673-0.
36. Zhang, X.W.; Tang, Z.F.; Lv, F.Z.; Pan, X.H. Excitation of axisymmetric and non-axisymmetric guided waves in
elastic hollow cylinders by magnetostrictive transducers. J. Zhejiang Univ.-Sci. A 2016, 17, 215–229. [CrossRef]
37. Yang, Q.; Li, Y.; Zhao, Z.; Zhu, L.; Luo, Y.; Zhu, J. Design of a 3-D rotational magnetic properties measurement
structure for soft magnetic materials. IEEE Trans. Appl. Supercond. 2014, 24, 1–4. [CrossRef]
218
applied
sciences
Article
Calculation of Guided Wave Dispersion
Characteristics Using a Three-Transducer
Measurement System
Borja Hernandez Crespo 1,2 , Charles R. P. Courtney 1, * and Bhavin Engineer 3
1 Department of Mechanical Engineering, University of Bath, Claverton Down, Bath BA2 7AY, UK;
bhc24@bath.ac.uk
2 TWI Ltd., Granta Park, Great Abington, Cambridge CB21 6AL, UK
3 Cranfield University, College Road, Cranfield, Bedfordshire MK43 0AL, UK;
bhavin.engineer@cranfield.ac.uk
* Correspondence: c.r.p.courtney@bath.ac.uk
Received: 3 July 2018; Accepted: 26 July 2018; Published: 29 July 2018
Featured Application: Rapid generation of dispersion curves for guided wave applications when
knowledge of the material properties and thickness of the structure to be inspected are unknown.
Abstract: Guided ultrasonic waves are of significant interest in the health monitoring of thin
structures, and dispersion curves are important tools in the deployment of any guided wave
application. Most methods of determining dispersion curves require accurate knowledge of the
material properties and thickness of the structure to be inspected, or extensive experimental tests.
This paper presents an experimental technique that allows rapid generation of dispersion curves for
guided wave applications when knowledge of the material properties and thickness of the structure
to be inspected are unknown. The technique uses a single source and measurements at two points,
making it experimentally simple. A formulation is presented that allows calculation of phase and
group velocities if the wavepacket propagation time and relative phase shift can be measured. The
methodology for determining the wavepacket propagation time and relative phase shift from the
acquired signals is described. The technique is validated using synthesized signals, finite element
model-generated signals and experimental signals from a 3 mm-thick aluminium plate. Accuracies to
within 1% are achieved in the experimental measurements.
Keywords: guided waves; lamb wave; dispersion curves; phase velocity; group velocity;
signal processing
1. Introduction
Guided waves have been extensively investigated in the recent years as a nondestructive testing
(NDT) technique because of their advantages compared to ultrasonic testing (UT) inspection. The key
benefit of this technology is the ability to interrogate the entire thickness of thin-walled structures over
large areas from a single location.
During guided wave propagation, waves experience dispersion that distorts the wave shape
as the wave propagates, due to a dependence of velocity on frequency [1,2]. Dispersion curves
show the relation of phase and group velocities against frequency, for a particular geometry and
material. Determination of dispersion curves is important for any guided wave application; accurate
dispersion curves enable wave modes of received wavepackets to be distinguished, specific wave
modes to be cancelled to clean the acquired signal, the propagation of wave modes in a particular

Appl. Sci. 2018, 8, 1253
direction using phased arrays [3] and the spatial location of damage in the structure to be determined
based on time-of-flight measurements [4]. Guided waves are being used in commercial products,
for instance to detect wall thickness losses due to corrosion for pipeline inspection in the oil and
gas industry. These commercial products generate a unique wave mode which propagates along
the structure avoiding the creation of multimodal excitations, as this would increase the complexity
of the analysis of the signals [5,6]. Problems with the application of guided waves occur when the
material properties or thickness of the structure to be inspected are unknown due to imprecise records,
commercial confidentiality or manufacturing and maintenance uncertainty. This is a particular problem
when evaluating composite structures, like wind turbine blades, where elastic constants are not often
available. To inspect this kind of structure in an industrial environment, it becomes impracticable as
currently there are no means of generating the dispersion curves for such situations. Therefore, there
is a need to be able to calculate in situ the phase and group velocities at frequencies of interest in a
quick and reliable way without requiring prior knowledge of material properties or thickness.
There has been much work published on the determination of dispersion curves describing
many analytical, numerical and experimental methods. For relatively simple structures like plates
or pipes, dispersion curves can be predicted using the commercially available software Disperse® ,
which uses the global matrix method [7]. Other methods, such as the transfer matrix method [8,9]
or pseudospectral collocation method [10], have been utilized to determine dispersion curves. Finite
element (FE) [11] and semi-analytical finite element [12,13] methods have been also used. All these
numerical and analytical methods require knowledge of material properties and thickness of the
structure to determine the dispersion curves.
Experimental methods do not suffer this limitation, since guided wave data is directly acquired
from the inspected structure. The most widely used experimental technique to measure dispersion
curves is the 2D fast Fourier transform (FFT) [14]. This signal processing technique requires the
acquisition of many signals along the wave propagation direction in order to carry out a double FFT in
time and space. The result is a wavenumber–frequency matrix, which can be rearranged to give the
phase velocity dispersion map. The acquisition of the signals from all required locations is manually
prohibitive; therefore, a laser scanning vibrometer (LSV) is commonly used to automatically obtain the
signals from preselected points. This device is highly sensitive and bulky, being restricted to controlled
areas like laboratories [15] and difficult to use in situ.
Other experimental techniques have been proposed recently. Harb and Yuan [16,17] presented a
noncontact technique using an air-coupled transducer (ACT) to generate the wave mode on the plate
and an LSV to acquire the propagating mode. The technique requires precise control of the incidence
angle of the ultrasonic pressure from the ACT upon the surface. By relating the frequency to the
incidence angle at which the wave amplitude is maximized, it is possible to calculate the phase velocity
using Snell’s law. However, the technique has limited applicability outside of the laboratory. In works
by Mažeika et al. [18,19], a zero-crossing technique is used to calculate the phase velocity based on the
measurement of the position of constant phase points of the pulse corresponding to zero amplitude
as a function of time. A similar approach is taken in the phase velocity method in [20], tracking the
peaks of the pulse rather than the zero crossings. To produce convincing results, both techniques
require the acquisition of a large number of signals at different distances, making these techniques
time-consuming. For the sake of reducing time and sampling points in space, Adams et al. [21]
presented a method using an array probe from which a time–space matrix can be created. A 2D filter is
applied to the matrix to extract the phase velocities. The technique was demonstrated to be valid in
simulation using FE analysis. However, for realistic probe sizes, the technique has limited experimental
applications. In [22], sparse wavenumber analysis was used to experimentally recover the dispersion
curves from an aluminum plate; however, seventeen sensors were needed to deploy the technique. An
alternative approach to getting dispersion curves from experimental signals is to use time–frequency
representations [23], where group velocity can be directly determined. However, the calculation of
phase velocity involves the integration of group velocity requiring precise values of ω and k at the
220
Appl. Sci. 2018, 8, 1253
lower limit, which are not easy to obtain experimentally. For medical purposes, two techniques, SVD
and LRT, were evaluated for extracting the dispersion curves of cortical bones in [24]. Both techniques
require a large number of receivers and transmitters.
Experimental calculation of bulk wave velocities in dispersive materials has been studied [25].
Sachse and Pao used the method of phase spectral analysis, where the phase delay is calculated using
the real and imaginary parts of the Fourier transform of the pulse, and then obtaining the group delay
through the differentiation of the phase characteristics of the pulse with respect to frequency. Pialucha
et al. presented the amplitude spectrum technique overcoming the requirement of the phase spectrum
technique for the separation of successive pulses in the time domain for bulk waves [26]. However, the
phase spectral technique provided better results when pulses can be separated.
In this paper, an experimental technique for straightforward calculations of phase and group
velocities of guided waves at frequencies of interest is presented. The technique requires just the
acquisition of two signals spaced a few cm using conventional transducers, enabling its application
on in-service structures located in poorly controlled environments. There is no requirement for
prior knowledge of the material properties or thickness of the inspected structure. In Section 2, the
formulation on which the experimental technique relies is presented. Section 3 describes in detail
the methodology to extract the group delay and phase shift from experimental signals. Successive
sections present validation tests using synthesized signals and simulated signals from FE analysis.
Finally, experimental signals from a 3 mm-thick aluminium plate are used to validate the technique as
a solution to the problem presented in the introduction.
2. Theoretical Basis
Assuming that the propagating pulse is sufficiently narrow-band that dispersion does not
significantly distort the wavepacket over the propagation distance, the signal can be approximated
as [27],
x
ψ( x, t) = G t − ei(kx−ωt) , (1)
vg
where x is propagation distance, t is time, ω = 2π f is the angular frequency of the harmonic wave
at frequency f and k = 2π/λ is the wavenumber. G (t) is a function defining the temporal pulse
shape. G (t) is defined such that the pulse has a peak at t = 0. v g is the group velocity at which the
packet propagates, and so if the trajectory of the wavepacket peak is defined as the point whereby

G t0 − x0 /v g = G (0), then
x0
vg = . (2)
t0
The wavepacket is shown schematically in Figure 1. The harmonic part of the signal has a phase of
φ = kx − ωt, (3)
and propagates at the phase velocity [6,27].

ω
vp = (4)
k
The phase of the harmonic signal at the centre of the wavepacket x0 , t0 is
φ0 = kx0 − ωt0 . (5)
Rearranging and substituting for v p and v g gives

x0 ω
φ0 = kt0 − , (6)
t0 k
221
Appl. Sci. 2018, 8, 1253

φ0 = kt0 v g − v p . (7)
For nondispersive waves, v p = v g and Equation (7) gives the expected result that the phase at the
wavepacket center is φ0 = 0 for all propagation times. Where the system is dispersive there is a finite
phase difference between the harmonic part of the signal and the wavepacket, φ0 , which increases
linearly with propagation time and is positive if v g > v p , negative if v g < v p . Assuming that this phase
difference can be measured experimentally, Equations (4) and (5) further yield two useful equations
for evaluating the phase velocity:
φ
x0 − k0
vp = , (8)
t0
x0
vp = φ
. (9)
t0 + ω0
In Figure 2, an illustrative example of a S0 -mode wavepacket at 150 kHz is shown. The solid
line is the S0 mode after 1.2-m propagation including the dispersion. The dashed line represents the
S0 mode after the propagation without dispersion, namely, the wavepacket was calculated using the
same phase velocity for all frequencies. In this example, the phase shift is negative (–1.26 radians),
which means the phase velocity is higher than the group velocity. Using (9), the calculation of the
phase velocity would be 1.2 m divided by the sum of 223 μs, which is the time of the wavepacket to
travel 1.2 m at group velocity, plus –1.34 μs which is the time that the signal at an angular frequency,
ω = 2π ·150000, takes to cover –1.26 radians. Group velocity and phase velocity are calculated using
(2) and (9), respectively: v g = 5381 ms−1 and v p = 5414 ms −1 .
‫ ݔ‬ൌ ‫ݔ‬଴ ǡ ߶ ൌ ߶଴
‫ݒ‬௚
‫ݐ‬଴
Figure 1. Wavepacket arriving at x0 at time t0 with phase φ0 propagating at group velocity v g . G (t) is
represented as the envelope of the wavepacket.
Figure 2. S0 -mode wavepacket of 5 cycles at 150 kHz after propagating 1.2 m, with dispersion and
without dispersion.
222
Appl. Sci. 2018, 8, 1253
3. Methodology
In this section, the methodology for determining the phase shift and hence the phase velocity is
presented. This general methodology is complemented by a description of a specific experiment setup
described in Section 6.1. The technique is based on the analysis of two signals acquired at different
distances in the direction of wave propagation. One of them will be treated as the baseline and the
other signal will be modified in time and phase in order to match with the baseline. The methodology
is composed of different signal processing techniques which are presented in a block diagram in
Figure 3.
Figure 3. Methodology’s block diagram to extract the optimum phase shift and time delay.
The experimental setup is composed of one transmitter and two receivers along the propagation
direction of the wave to be measured, as shown in Figure 4. For isotropic systems the wave velocity is
independent of direction. The technique analyses the effect of propagation between the first receiver
and the second receiver. Spacing between receivers of a few centimeters is selected in order to avoid
large phase shifts and waveform deformations due to dispersion.
7UDQVPLWWHU 5HFHLYHUV
Figure 4. Schematic of the test setup.
The first step in analyzing the data is the isolation of the desired wavepacket by removing the
rest of the signal. The boundaries of each wavepacket are determined by studying the slope of its
envelope. The boundaries are set at the closest zero slopes before and after each peak, see example of
real signals in Figure 5. Once the region containing the wavepacket is identified, the remainder of the
signal is discarded. Both signals are truncated with the same boundaries in order to maintain the time
difference between the wavepackets, w1 and w2 . The boundaries used are the left limit of the signal of
the closest receiver to the transmitter and the second limit of the signal of the further receiver, as can
be seen in Figure 6. Afterwards, the Hilbert Transform (H( . . . )) [28] is applied to both wavepackets
and the magnitude extracted in order to determine the envelope of each wavepacket:
e1 (t) = | H (w1 (t))|, (10)
e2 (t) = | H (w2 (t))|. (11)
223
Appl. Sci. 2018, 8, 1253
D
E
Figure 5. Wavepacket detection and boundary determination established by the algorithm for two
signals acquired at: (a) 30 cm and (b) 35 cm from the transmitter using a d33 -type transducer.
D
E
Figure 6. Shortening of the signals to reduce computational time. The same limits are used to truncate
each signal in order to maintain the time difference between wavepackets. Signals acquired at: (a) 30 cm
and (b) 35 cm from the transmitter.
These two envelopes are then cross-correlated to determine the time delay between them:
Δt = arg max ((e1 e2 )(τ )), (12)

τ
where τ is the applied time delay and Δt the value that maximizes the correlation. Once this time delay
is known, the phase delay to the harmonic part of w2 (t) that maximizes the cross-correlation between
the delayed phase-shifted wavepacket w2∗ (t − Δt) and w1 (t) is calculated. This process consists of the
modification in phase of the signal from the furthest receiver by a phase angle, φ, and finding the
phase shift that maximizes the cross correlation. The phase shift, φ, is applied to w2∗ ; then, the cross
correlation between w1 and w2∗ is calculated. The optimum phase shift is the phase angle when the
correlation coefficient with time delay, Δt, is highest.
φopt = arg max ((w1 w2∗ )(Δt)(φ)) (13)

φ
224
Appl. Sci. 2018, 8, 1253
This methodology enables the calculation of the lag time and phase shift between the two signals
acquired at different distances. By knowing these three variables (distance, time and phase) and also
the excitation frequency, (2) and (9) can be used to determine the group and phase velocity, respectively.
The proposed methodology in this paper is based on the acquisition of two signals spaced a few
centimeters apart, since for such short distances the distortion for any wave mode is relatively small,
enabling a straightforward determination of the phase shift.
4. Synthetic Signal Analysis

The first validation was carried out using synthesized signals. The signal synthesis is based on
the adjustment in frequency and wavenumber of the input signal in the frequency domain. Knowing
the waveform at one point, in this case the input signal at the transmitting point, and the dispersion
characteristics of each wave mode, the wave modes can be reconstructed in the time domain at any
distance [29]. If u( x, t) is the reconstructed wave mode at a distance x from the transmitting point, t is
the time, F (ω ) is the Fourier transform of the input signal and k(ω ) is the wavenumber as a function
of the angular frequency, then
' ∞
u( x, t) = F (ω )ei(k(ω ) x−ωt) dω (14)
−∞
The relation between wavenumber and angular frequency is extracted directly from the dispersion
curves provided by Disperse® and introduced in the integral.
As an initial test of the signal processing method, developed signals are synthesized that are
highly pure, without noise and overlapping between modes. The higher the sampling frequency is set,
the better accuracy the technique provides. The sampling frequency was set at 10 MHz, since it has
been observed that it provides a good balance between computational time and accuracy.
A set of signals for symmetric (S0 ), antisymmetric (A0 ) and shear horizontal (SH0 ) wave modes
were created by evaluating equation (14) in MATLAB for two propagation distances, 30 cm and 35 cm:
the analysed propagated distance (Δd) in this case is 5 cm. The analysed frequencies are from 60 to
370 kHz with steps of 10 kHz.
The results after applying the signal processing algorithm described in Section 3 to the synthesized
signals are presented in Figure 7. The results (black circles) accurately reconstruct the theoretical values
from Disperse® (grey lines). In Figure 8, the extracted phase shift has been represented against the
frequency for the three fundamental modes. For the S0 mode, the phase values have negative values
decaying from –2 degrees at 60 kHz to –52 degrees at 370 kHz. For the A0 mode, conversely, the phase
is positive, increasing along the frequency, since the phase velocity is lower than the group velocity;
from 377 degrees at 60 kHz to 673 degrees at 370 kHz. In this case, the phase values are much higher
than the S0 and SH0 ones, because the A0 mode is highly dispersive at these frequencies. For the
SH0 mode, the extracted phase values are 0 degrees at 60 kHz and 1 degree at 370 kHz. The SH0 is a
nondispersive mode so the phase value remains at near 0 degrees.
225
Appl. Sci. 2018, 8, 1253
s a none so
6 6+
$
6
6+
$
Figure 7. Comparison of the phase velocity and group velocity of S0 , A0 and SH0 between the results
from the synthesized signals (black circles) and theoretical values (grey lines).
$
6+
6
Figure 8. Comparison of the phase shift of S0 , A0 and SH0 between the results from the synthesized
signals (black circles) and theoretical values (grey lines).
The sampling frequency is an important factor, since it determines the degree of resolution of
the group velocity. As these examples are performed for short propagation distances, the time that
the wave modes take to cover that distance is small. Thus, low sampling frequencies are not able to
determine the group velocity accurately. Faster modes require higher sampling frequencies to get the
same resolution as slower modes. In this case of 5 cm spacing, the propagation time for the S0 mode at
200 kHz is 9.3 μs; a sampling frequency of 1 MHz will have a time resolution of 1 μs which yields a
velocity resolution for this particular example of 556 m/s; with a sampling frequency of 10 MHz, the
velocity resolution will be 58.4 m/s; and with a sampling frequency of 100 MHz, the velocity resolution
will be 5.8 m/s. This resolution issue can be observed in Figure 7; at lower velocities the curves are
smoother, and at higher velocities (S0 mode) the curves have poorer velocity resolution.
226
Appl. Sci. 2018, 8, 1253
5. Finite Element Analysis

A three-dimensional FE analysis was performed in Abaqus to simulate signals. While the previous
analysis required the input of the dispersion curves to create the propagated signals, this FE analysis
does not require prior knowledge of the dispersion characteristics.
FE models were created simulating guided wave propagation at five different frequencies in an
aluminium plate (1000 × 1000 × 3 mm). The transmitter and receivers were modelled as ideal point
transducers. The transmitting point was placed at the centre of the plate, where the coordinate system
was established. A force was applied in the x-direction to simulate a shear transducer. This generated
S0 and A0 modes along the x-axis and the SH0 mode along the y-axis. The input signal was a 5-cycle
sine with a Hanning window at central frequencies of 60, 80, 100, 120 and 140 kHz. Multiple receiver
points were located on the positive x-axis and positive y-axis from the centre of the plate to the edge
every 1 cm, resulting in 50 receivers at each axis. The time and spatial discretization for the finite
element model was established at 10-2 μs and 1 mm, respectively, in order to ensure convergence and
to obey the Nyquist sample-rate criterion. The mesh was formed by C3D8 elements.
The FE analysis software provides the results decoupled for each axis direction, so it is
straightforward to evaluate each wave mode separately. In Figure 9, the results have been computed in
cylindrical coordinates to show clearly the three wave modes of propagation at each axis. In Figure 9a,
showing radial displacement, S0 and A0 modes are depicted, S0 being the faster mode. In Figure 9b
showing tangential displacement, the SH0 mode propagates along the y-axis, and in Figure 9c, showing
out-of-plane displacement, the A0 mode is clearly present.
6
D
$
E F
$
6+
Figure 9. Images of the finite elementsimulation of the wave propagation in polar-coordinate

instantaneous displacement is shown 76 μs after excitation. (a) Radial displacement, (b) tangential
displacement and (c) out-of-plane displacement. Scale bars in meters.
Since the plate is relatively small, overlaps between wave modes and echoes from edges are
produced, as can be seen in Figure 10. This is a particular problem at lower frequencies (60 and 80 kHz)
due to the length of the pulses. Therefore, an additional optimisation step has been added to minimize
the error introduced by the overlapping. The FE model has 50 receiver points, every 1 cm along
the axis so the five least-overlapped signals are selected for analysis. The selection is performed by
analyzing the amplitude at the beginning and end of the envelope of the wavepacket of interest; the
227
Appl. Sci. 2018, 8, 1253
signals with the lowest amplitude at those regions are selected. Then, the signal processing described
in Section 3 is applied for all the pair combinations between the five selected receivers. Once the
phase and group velocities are calculated for each combination, the velocities are averaged to give
the final velocity estimates. Any outliers that are introduced during selection of the five signals are
eliminated. As shown in Figure 11, the phase velocity matches very well with the theoretical values
from Disperse® . In the group velocity graph, the results correlate well with the theoretical velocities,
having a noticeable variation in the S0 value at 60 kHz. This variation occurs at 60 kHz due to the
overlap between the S0 mode and the A0 mode, which is more pronounced for the longer pulses at
this frequency. This error does not appear for the A0 mode because the signals used for the analysis
were out-of-plane displacements, where the S0 mode is practically imperceptible. Note that the results
improve as the frequency increases, since the wavelength decreases and mode separation increases.
Overlapping of pulses is an issue for this technique, as it changes the phase and waveform of the
analyzed mode, highly distorting the results and impeding its application.
Figure 10. Signal acquired at 33 cm from excitation point at 60 kHz.
Figure 11. Comparison of the phase velocity and group velocity of S0 , A0 and SH0 between the results
from the synthesized signals (black circles) and theoretical values (grey lines).
The results for the phase velocity are better than for the group velocity, more notably when
overlapping occurs. This is because of the phase-shift term in the denominator in (9). This contribution
in the phase velocity equation minimizes the erroneous calculated value of the propagation time and
228
Appl. Sci. 2018, 8, 1253
also provides more resolution on the calculation of the phase velocity, removing the stepped shape
seen in the group velocity.
6. Experimental Demonstration
6.1. Test Setup

The specimen is a 4 m by 2 m aluminium plate of 3 mm thickness. Two piezoelectric transducers
were used, one transmitter and one receiver. The transmitter was fixed at the centre of the plate using
a tool that provides force over the transducer; and the receiver was manually attached at various
distances from the transmitter. For the evaluation of the S0 mode, the receiver was placed at 45, 50 and
55 cm from the transmitter. For the case of the A0 and SH0 modes, the receiver was set at 25, 30 and
35 cm from the transmitter. Those distances were selected to minimize the wavepacket distortion due
to overlaps.
A pulser-receiver (Teletest® unit) was used to excite the transducer and acquire the signals. Two
independent channels were used for the excitation and acquisition. A PC with the Teletest® software
was used to control the pulser-receiver and configure the experimental parameters. Input signals of
five-cycle Hanning-windowed bursts with centre frequencies from 40 kHz to 140 kHz in steps of 5 kHz
were used to excite the transmitter. This frequency range was chosen since it is the operating frequency
range of the transducers used in the experiment. Signals were recorded with a 1 MHz sampling
frequency; 32 averages were taken, and the record length of the signals was 1 ms. The experimental
setup is presented in Figure 12.
&RPSXWHU
6+ZDYH
ZDYH
SURSDJDWLRQ 6$ZDYH
ZDYH
SURSDJDWLRQ
3XOVHU5HFHLYHU 7UDQVPLWWHU
5HFHLYHU
Figure 12. Experiment setup: Left–diagram of the experiment, Right–photo of the real
experiment environment.
Two types of piezoelectric transducers (PI Ceramic GmbH) were used in the experiment, a shear
piezoelectric element (d15 type) and a compressional piezoelectric element (d33 type), both with an
active contact area of 13 mm by 3 mm. Figure 13 shows the direction of poling and corresponding
shear deformation for the shear transducer. The shear transducer was used to evaluate the S0 and SH0
wave modes, and the A0 wave mode was analyzed using the compressional transducer, which mainly
produces an out-of-plane vibration. Using the shear transducer, S0 waves are generated in the direction
of the poling axis, and SH0 in the perpendicular axis. A small amount of A0 is also generated on the
poling axis. In the case of the compressional transducer, S0 and A0 are created omnidirectionally. The
averaged amplitudes of the S0 and SH0 modes using the shear transducer at the propagation distances
specified above are 13 mV and 33 mV, respectively; and the amplitude of the A0 mode is 63 mV using
the compressional transducer.
229
Appl. Sci. 2018, 8, 1253
Figure 13. Schematics of the shear transducer. (a) Undeformed, (b) deformed.
Experimental signals were acquired at a sampling frequency of 1 MHz providing a poor time
resolution for the wave velocities of interest. Therefore, the sampling rate of the signals was increased
by a factor of 10 using cubic spline interpolation to give a time accuracy of 0.1 μm (10 MHz). The
interpolation is a necessary step when the propagation distance is a few cm, as the propagation time
of the faster modes, such as S0 , is of the same order of magnitude as the original sampling period.
By reducing the sampling period by means of interpolation, the accuracy of the group velocity is
increased. However, a small error is introduced in the signal due to the interpolation. Overall, the
improvements gained for the velocity calculation are more significant than the errors introduced in
the signals. The signal-to-noise ratios of the S0 , A0 and SH0 signals were around 30 dB, 37 dB and
35 dB, respectively.

The results presented in Figures 14 and 15 have been extracted from signals at two different
spacings: 5 cm and 10 cm. The velocities extracted from the experimental signals correlate quite well
with the theoretical velocities. However, there are slight variations around the theoretical velocities.
These variations can be caused by a number of factors, such as the aforementioned interpolation
and overlapping, but also by the manipulation of the transducers, which in this case were handled
manually. Small variations in the correct spacing distance between transducers lead to errors in the
calculation of the velocities. The velocity–distance relationship is linear so the relative error in the
calculated velocity due to misplacement is the same as the relative distance error. An error of 1 mm in
an intended 5 cm spacing distance will cause a 2% error in the velocity or an error of 110 ms−1 in the
S0 mode (≈ 5400 m/s), for example. On the other hand, with a 1 mm error transducer placement for a
10 cm spacing, the error will be of 1% of the velocity. Therefore, large errors can occur where small
propagation distances are used. A0 and SH0 exhibit less absolute error variation than S0 due to their
slower velocities, the generated errors being proportional to the velocity.
In this paper, two propagation distances have been chosen: 5 and 10 cm. From the S0 mode in
Figure 15, it can be seen that the calculated velocities at 10 cm are more accurate, and the variance
is also smaller. This is for two reasons: the first is the reduction in the error caused by an incorrect
transducer placement, and the second is the implicit increase of the propagation time, which reduces
the error in determining that time. Other factors that can introduce errors in the velocity calculation
are the variations of the boundary conditions of the transducer, such as the applied pressure over the
transducer at each receiving location, as well as the difference of transfer function of each transducer.
This problem was overcome utilizing the same transducer to acquire the signals at all locations.
Similarly, a change in temperature between measurements will cause velocity changes leading to
erroneous values. This can be avoided by acquiring results in quick succession to minimize the impact
of the temperature in the results. Using three transducers, one transmitter and two receivers would
avoid the issues relating to moving the transducers.
230
Appl. Sci. 2018, 8, 1253
Figure 14. Phase velocity and group velocity created from experimental signals for 5 cm propagation
distance (black circles) and theoretical values (grey lines).
Figure 15. Phase velocity and group velocity created from experimental signals for 10 cm propagation
distance (black circles) and theoretical values (grey lines).
231
Appl. Sci. 2018, 8, 1253
7. Conclusions
This paper has presented a new methodology for creating dispersion curves based on measuring
the phase difference and time lag between two pulses acquired at different distances. Firstly, the
theoretical basis of the method has been presented where the phase velocity has been related to the
phase at the pulse centre for systems of limited dispersion; then, a signal processing methodology to
extract the phase shift and time lag between two signals and subsequently calculate the phase and
group velocities has been outlined. Three different tests were performed to validate the formulation
and methodology; using synthesized signals, signals from an FE model and experimental signals. In
the three cases the performance achieved agrees well with the theoretical values.
There are some limitations to the proposed method for the extraction of the phase shift to
calculate the dispersion curves. Highly dispersive modes, like high-order modes, are more challenging
to evaluate; to mitigate this issue, much shorter spacing distances between transducers should be
established to minimize the mode-shape distortion. Overlapping between wavepackets limits the
method’s applicability, since it distorts the apparent phase of the mode being analyzed. Hence,
relatively large samples are required to avoid overlapping between wavepackets and edge echoes.
The size of the sample can be smaller at higher frequencies. The velocity–frequency spectrum that
can be evaluated using this method will be determined by the excitation mechanism used to generate
the guided waves in the solid media. With appropriate experimental design, errors of less than 1%
are obtained.
For future work, the proposed technique will be applied to more complex structures, such as
composite plates where the elastic constants are not easy to obtain. In this case, by studying the
phase and group velocities as a function of propagation direction, the full angular dependency of the
dispersion curves will be obtained.
Author Contributions: B.H.C. conducted the modeling, analysis, experiments and drafted the manuscript.
C.R.P.C. drafted the theory and helped edit the manuscript. B.E. helped edit the manuscript.
Funding: This research was funded by the European Commission under the Marie Curie project grant number
FP7-PEOPLE-2012 ITN 309395 “MARE-WINT”.
References
1. Rose, J.L. Ultrasonic Waves in Solid Media; Cambridge University Press: Cambridge, UK, 1999.
3. Wang, W.; Zhang, H.; Lynch, J.P.; Cesnik, C.E.; Li, H. Experimental and numerical validation of guided
wave phased arrays integrated within standard data acquisition systems for structural health monitoring.
Struct. Control Health Monit. 2018, 25, e2171. [CrossRef]
4. Wang, W.; Bao, Y.; Zhou, W.; Li, H. Sparse representation for Lamb-wave-based damage detection using a
dictionary algorithm. Ultrasonics 2018, 87, 48–58. [CrossRef] [PubMed]
5. Cawley, P.; Lowe, M.; Alleyne, D.; Pavlakovic, B.; Wilcox, P. Practical long range guided wave
inspection-applications to pipes and rail. Mater. Eval. 2003, 61, 66–74.
6. Rose, J.L. Ultrasonic Guided Waves in Solid Media; Cambridge University Press: Cambridge, UK, 2014.
7. Pavlakovic, B.; Lowe, M.; Alleyne, D.; Cawley, P. Disperse: A general purpose program for creating dispersion
curves. In Review of Progress in Quantitative Nondestructive Evaluation; Springer: Boston, MA, USA, 1997;
pp. 185–192.
8. Lowe, M.J. Matrix techniques for modeling ultrasonic waves in multilayered media. IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 1995, 42, 525–542. [CrossRef]
9. Nayfeh, A.H. The general problem of elastic wave propagation in multilayered anisotropic media. J. Acoust.
Soc. Am. 1991, 89, 1521–1531. [CrossRef]
232
Appl. Sci. 2018, 8, 1253
10. Quintanilla, F.H.; Fan, Z.; Lowe, M.; Craster, R. Dispersion loci of guided waves in viscoelastic composites of
general anisotropy. In AIP Conference Proceedings; AIP Publishing: Melville, NY, USA, 2016; Volume 1706,
p. 120014.
11. Yu, X.; Ratassepp, M.; Fan, Z. Damage detection in quasi-isotropic composite bends using ultrasonic feature
guided waves. Compos. Sci. Technol. 2017, 141, 120–129. [CrossRef]
12. Hayashi, T.; Song, W.-J.; Rose, J.L. Guided wave dispersion curves for a bar with an arbitrary cross-section, a
rod and rail example. Ultrasonics 2003, 41, 175–183. [CrossRef]
13. Cong, M.; Wu, X.; Liu, R. Dispersion analysis of guided waves in the finned tube using the semi-analytical
finite element method. J. Sound Vib. 2017, 401, 114–126. [CrossRef]
14. Alleyne, D.; Cawley, P. A two-dimensional Fourier transform method for the measurement of propagating
multimode signals. J. Acoust. Soc. Am. 1991, 89, 1159–1168. [CrossRef]
15. Ostachowicz, W.; Güemes, A. New Trends in Structural Health Monitoring; Springer Science & Business Media:
London, UK, 2013; Volume 542.
16. Harb, M.; Yuan, F. A rapid, fully non-contact, hybrid system for generating Lamb wave dispersion curves.
17. Harb, M.; Yuan, F. Non-contact ultrasonic technique for Lamb wave characterization in composite plates.
18. Mažeika, L.; Draudvilienė, L. Analysis of the zero-crossing technique in relation to measurements of phase
velocities of the Lamb waves. Ultragarsas 2010, 65, 7–12.
19. Draudvilienė, L.; Aider, H.A.; Tumsys, O.; Mažeika, L. The Lamb waves phase velocity dispersion evaluation
using an hybrid measurement technique. Compos. Struct. 2018, 184, 1156–1164. [CrossRef]
20. Moreno, E.; Galarza, N.; Rubio, B.; Otero, J.A. Phase Velocity Method for Guided Wave Measurements in
Composite Plates. Phys. Procedia 2015, 63, 54–60. [CrossRef]
21. Adams, C.; Harput, S.; Cowell, D.; Freear, S. A phase velocity filter for the measurement of Lamb wave
dispersion. In Proceedings of the 2016 IEEE International Ultrasonics Symposium (IUS), Tours, France, 18–21
September 2016; pp. 1–4.
22. Harley, J.B.; Moura, J.M. Sparse recovery of the multimodal and dispersive characteristics of Lamb waves.
J. Acoust. Soc. Am. 2013, 133, 2732–2745. [CrossRef] [PubMed]
23. Latif, R.; Aassif, E.; Maze, G.; Moudden, A.; Faiz, B. Determination of the group and phase velocities from
time-frequency representation of Wigner-Ville. NDT E Int. 1999, 32, 415–422. [CrossRef]
24. Xu, K.; Ta, D.; Cassereau, D.; Hu, B.; Wang, W.; Laugier, P.; Minonzio, J.-G. Multichannel processing for
dispersion curves extraction of ultrasonic axial-transmission signals: Comparisons and case studies. J. Acoust.
Soc. Am. 2016, 140, 1758–1770. [CrossRef] [PubMed]
25. Sachse, W.; Pao, Y.-H. On the determination of phase and group velocities of dispersive waves in solids.
J. Appl. Phys. 1978, 49, 4320–4327. [CrossRef]
26. Pialucha, T.; Guyott, C.; Cawley, P. Amplitude spectrum method for the measurement of phase velocity.
27. Graff, K.F. Wave Motion in Elastic Solids; Courier Corporation: Chelmsford, MA, USA, 2012; pp. 62–65.
28. Michaels, J.E.; Lee, S.J.; Croxford, A.J.; Wilcox, P.D. Chirp excitation of ultrasonic guided waves. Ultrasonics
2013, 53, 265–270. [CrossRef] [PubMed]
233
applied
sciences
Article
Damage Imaging in Lamb Wave-Based Inspection of
Adhesive Joints
Magdalena Rucka *, Erwin Wojtczak and Jacek Lachowicz
Department of Mechanics of Materials and Structures, Faculty of Civil and Environmental Engineering, Gdansk
University of Technology, Narutowicza 11/12, 80-233 Gdansk, Poland; erwin.wojtczak@pg.edu.pl (E.W.);
jacek.lachowicz@pg.edu.pl (J.L.)
* Correspondence: magdalena.rucka@pg.edu.pl or mrucka@pg.edu.pl; Tel.: +48-58-347-2497
Received: 5 February 2018; Accepted: 27 March 2018; Published: 29 March 2018
Abstract: Adhesive bonding has become increasingly important in many industries. Non-destructive
inspection of adhesive joints is essential for the condition assessment and maintenance of a structure
containing such joints. The aim of this paper was the experimental investigation of the damage
identification of a single lap adhesive joint of metal plate-like structures. Nine joints with different
defects in the form of partial debonding were considered. The inspection was based on ultrasonic
guided wave propagation. The Lamb waves were excited at one point of the analyzed specimen by
means of a piezoelectric actuator, while the guided wave field was measured with the use of a laser
vibrometer. For damage imaging, the recorded out-of-plane vibrations were processed by means of
the weighted root mean square (WRMS). The influence of different WRMS parameters (i.e., the time
window and weighting factor), as well as excitation frequencies, were analyzed using statistical
analysis. The results showed that two-dimensional representations of WRMS values allowed for the
identification of the presence of actual defects in the adhesive film and determined their geometry.
Keywords: adhesive joint; single lap joint; non-destructive testing; damage identification; Lamb
waves; scanning laser vibrometry
1. Introduction
Adhesive joints of metal elements are a kind of non-separable connections with numerous
applications, e.g., in the aviation, machine, or automotive industry [1]. The strength and durability
of adhesive joints are essential for the safety of engineering structures. They are highly susceptible
to any defects resulting from mistakes that were committed during the preparation and joining of
elements, e.g., poor condition of the adherend surface or a lack of glue on a part of the overlay.
Additionally, a gradual deterioration in the quality of the adhesive material cannot be neglected.
Typical adhesive joints are not capable of being inspected based on visual techniques as the place
of damage occurrence (the adhesive film) is hidden between the joined parts. Destructive testing
provides very accurate results [1]. This approach allows for the determination of the influence of certain
parameters (e.g., damage existence [2,3] or adhesion condition [4]) on the strength of representative
samples, which were usually prepared in laboratory conditions and to predict the state of existing joints.
Destructive testing is not free of disadvantages, it is, in general, a high-cost solution and cannot usually
be applied to existing structures. Therefore, different non-destructive testing (NDT) methods have
been adopted for the inspection of adhesive joints [5] including ultrasonics [6,7], thermography [8],
radiography [7], digital image correlation [9,10], or electromechanical impedance spectroscopy [11].
In recent years, one of the most promising techniques for damage assessment is the use of guided
wave propagation (e.g., [12–14]). Lamb waves are a specific type of guided wave that occur in a media
that is restrained by two parallel surfaces, e.g., thin plate-like structures. The ability of Lamb waves to
interact with any kind of defect makes them particularly useful in the diagnostics of various elements

Appl. Sci. 2018, 8, 522
of engineering structures, including adhesive joints. In previous studies, guided waves have been
proven to be an effective technique for the inspection of adhesive lap joints [15–18], adhesive bonds
between composites [19–21], adhesive butt joins of plates [22], or the detectability of the adhesion
level [23].
In the guided wave propagation technique, the wave is usually excited at one point of a
structure and is collected at a number of points that are spread over the whole structure. In general,
two measurement strategies can be used. The first method depends on using a small number
of ultrasonic or piezoelectric transducers, followed by an analysis of registered time histories.
This strategy is mainly used for structural health monitoring (SHM). Another approach is directed
towards non-contact techniques, such as scanning laser Doppler vibrometry (SLDV) and the sensing of
the full guided wave field (e.g., [24–26]). This strategy is more appropriate for quality assurance testing
and periodic nondestructive inspection. Particular signals that are registered at numerous points
distributed over a monitored area provide a two-dimensional representation of propagating waves at
specific time instances. If any damage exists in an analyzed structure, a disturbance in the wavefront
shape can be observed. The perturbation is more or less clearly visible depending on the type, size, and
position of the defect. This visual assessment of SLDV maps allows for one to determine the presence
of damage, but does not give a clear image due to the multiple reflections of waves from discontinuities
and element edges [25]. The determination of the actual shape and size of the existing defect requires
additional processing and damage imaging [27–29]. One of the simplest but most efficient techniques
is based on vibration energy distribution and utilizes the calculation of a root mean square (RMS).
This method was improved to be more applicable for damage detection by defining the weighted root
mean square (WRMS) method [30]. Recently, the RMS has been successfully used in various damage
detection applications [25,26,31–36]. Żak et al. [30] investigated aluminum and composite plates as
well as composite elements of a helicopter, all with the damage being simulated by an additional
mass. They concluded that better resolution in WRMS maps corresponded to higher values of the
weighting factor (linear or square). Localization of damaged rivets in a stiffened plate structure based
on RMS maps was conducted by Radzieński et al. [31] where the authors used a five-cycle burst of
different frequencies (5 kHz, 35 kHz, and 100 kHz). The best results were obtained in the case of
an excitation signal of 100 kHz, where the wavelength was shorter than the distance between two
rivets. Lee and Park [32] considered an aluminum plate with notches of various orientation (tangential,
perpendicular, and diagonal to the incident wave front). Based on RMS images, only tangential damage
was successfully detected without a comparison to the intact images. Saravanan et al. [33] used radially
weighted and factored RMS for damage localization in an aluminum plate. Detection of delamination
in a composite T-joint using RMS images was studied by Geetha et al. [34]. Marks et al. [35] presented
experimental investigations on the detection of disbonds in a stiffened panel based on the RMS plots.
Wave packets with three different frequencies, namely 100 kHz, 250 kHz, and 300 kHz, were applied.
The authors concluded that the excitation frequency is an important consideration for the inspection of
disbonds. In the conducted works, only higher frequencies (250 kHz and 300 kHz) were able to detect
damage. Lee et al. [36] applied the WRMS to localize the notches and corrosion areas in aluminum
plates and chose 2 as the weighting parameter. Kudela et al. [25] studied the detection of delamination
in a composite plate using WRMS with the weighting factor set as 2.7. They also proposed a selective
WRMS that enabled the elimination of edge reflections. Detection of disbond inserts in composite
elements was performed by Pieczonka et al. [26] using RMS and three-dimensional (3D) scanning.
In the conducted experimental investigations, it was noticed that in-plane RMS maps provided a better
indication of damage than out-of-plane RMS maps. Summarizing all of the works that are reported
above, the main problem that is connected with damage imaging using WRMS is the fact that the
weighting factor is set arbitrary with no clear explanation for the choice of a specific value.
This study presents an approach to the damage identification of adhesively bonded joints that is
based on the Lamb wave propagation. The guided wave field was measured with the use of scanning
laser Doppler vibrometry and then processed by means of the weighted root mean square. The object of
236
Appl. Sci. 2018, 8, 522
the research was a single lap adhesive joint of metal plate-like structures. To verify the feasibility of the
proposed approach, nine specimens with defects that were differing in size and position were tested.
The influences of different WRMS parameters (i.e., the time window and weighting factor), as well
as excitation frequencies, were analyzed. A novel element in this research was the application of the
statistical analysis in determining the WRMS parameters. It has been proven that the obtained WRMS
maps of damaged joints were characterized by bimodal distributions that created the opportunity to
determine the optimal value of the time window, weighting factor, and excitation frequency.
2.1. Description of Specimens

The object of the research was a single lap adhesive joint of steel plates. Dimensions of each
adherend were 270 mm × 120 mm × 3 mm. The length of the overlap was 60 mm (the half-width
of the plate). The geometry of the analyzed joint is shown in Figure 1. The plates were degreased
with Loctite-7063 cleaner and dried. Then, the surface of the overlap was treated with abrasive paper
(type P120) and was degreased again just before applying the glue. The adhesive layer was prepared
with the use of the epoxy-based glue Loctite EA 9461. The mean value of the bondline thickness for all
of the specimens was equal to 0.216 mm and the standard deviation had a value of 0.043 mm, which
resulted in the coefficient of variation being equal to 0.198.
Figure 1. Geometry of the analyzed single lap joint: (a) plan view; and, (b) side view.
Investigations were conducted for nine types of specimens (Figure 2): one intact (#1) and eight
(#2–#9) with different defects in the form of partial debonding. The defects were modelled as a lack
of glue on a part of the overlap surface. A PTFE tape of 0.2 mm thickness was used to prevent the
glue from leaking to the debonding areas of the damaged joints. The adherends of each specimen
were separated after the experimental measurements to determine the exact geometry of defects in the
prepared joints. Photographs of the disconnected joints are presented in Figure 3.
Figure 2. Investigated specimens: intact (#1) and with defects (#2–#9).
237
Appl. Sci. 2018, 8, 522
Figure 3. Photographs of the investigated specimens after separation (regions outlined with yellow
line denote kissing defects).
It is visible in Figure 3 that the defects had slightly different shapes and areas than was assumed.
The glue that leaked onto the surfaces covered with the PTFE tape (i.e., intended voids) did not affect
the WRMS results because the tape prevented the plates from connecting even though there were
some adhesive leakages. However, in addition to the intended voids, some additional areas resulting
from the improper adhesion of the adhesive to the adherend (so-called kissing defects or low volume
defects [8]) were identified for specimens #3 to #9. The kissing defect regions are outlined in Figure 3
by the yellow line. To quantify the real defect areas, including both the intended voids and kissing
defects, the photographs seen in Figure 3 were processed using the ImageJ program [37]. The results
are given in Table 1.
Table 1. Designed and determined area of defects.
Determined Surface of Determined

Design Surface of Design Relative
No. of Sample Defect Area Including Relative
Defect Area (cm2 ) Defect Surface (%)
Kissing Defects (cm2 ) Defect Surface (%)
#1 - - - -
#2 24.00 24.00 33.33% 33.33%
#3 48.00 49.11 66.67% 68.21%
#4 15.00 17.64 20.83% 24.50%
#5 24.00 26.03 33.33% 36.16%
#6 24.00 26.33 33.33% 36.57%
#7 24.00 29.55 33.33% 41.05%
#8 8.00 9.09 11.11% 12.63%
#9 8.00 13.34 11.11% 18.53%

Experimental measurements consisted of the excitation of Lamb waves in the lower adherend
and the collection of wave propagation signals in the selected area of the specimen. The experimental
setup is presented in Figure 4. The excitation signal was generated by means of the arbitrary function
generator AFG 3022 (Tektronix, Inc., Beaverton, OR, USA) and the high-voltage amplifier PPA 2000
238
Appl. Sci. 2018, 8, 522
(EC Electronics, Krakow, Poland). Lamb waves were excited by the plate piezoelectric actuator
NAC2024 (Noliac, Kvistgaard, Denmark), with dimensions of 3 × 3 × 3 mm3 and a resonant frequency
of over 500 kHz. The actuator was bonded to the bottom side of the specimen (Figure 4b) using
petro wax 080A109 (PCB Piezotronics, Inc., Depew, NY, USA). The excitation signal was a five-peak
wave packet obtained from a sinusoidal burst with a carrier frequency of 200 kHz modulated by
the Hanning window. There were also additional excitation signals for specimens #3 and #6 that
differed in the frequency value from the basic variant (100 kHz and 50 kHz, instead of 200 kHz).
Acquisition of the wave propagation signals was realized by the scanning head of the laser vibrometer
PSV-3D-400-M (Polytec GmbH, Berlin, Germany) with a VD-09 velocity decoder (Figure 4a). Time
signals were acquired with a sampling frequency of 2.56 MHz and were averaged 40 times to improve
the signal-to-noise ratio. The excitation signal was triggered as external and monitored using a
reference channel of the vibrometer. To improve the backscatter of light, the scanned surface was
covered with retro-reflective sheeting. Velocity values (out-of-plane components) were recorded in the
time domain in 7381 points that were distributed at the top side of the specimen (on the surface of the
upper plate with no PTFE tape attached, see Figure 3). The area was scanned point by point in the
mesh of 61 rows and 121 columns (Figure 5), giving a scan resolution of 1.933 mm and about nine scan
points per the shortest wavelength. Signals sensed at all of the points allowed for us to prepare a map
of the propagating waves, i.e., a two-dimensional plane view at a specific time instance.
Figure 4. Experimental setup: (a) instrumentation for generation and acquisition of Lamb waves; and,
(b) detail showing the piezo actuator attached to the specimen.
Figure 5. Investigated specimen with indicated scanning points.
2.3. Weighted Root Mean Square Damage Imaging

The guided wave field that was acquired with the use of SLDV can be further processed to more
accurately identify the position and shape of defects. This processing technique depends on the
239
Appl. Sci. 2018, 8, 522
calculation of the weighted root mean square for each registered time signal and the creation of a
damage map. The WRMS for a time signal s(t) can be defined as follows [25,26,30,33,36]
"
' t2
1
WRMS = w(t)s(t)2 dt (1)
t2 − t1 t1
where w(t) denotes a weighting factor and (t2 − t1 ) is length of the time window with t1 denoting
the beginning of the time window and t2 denoting the end of the time window. For a discrete signal
sr = s(tr ) sampled with Δt interval and calculated from the interval T = ( N − 1)Δt = t2 − t1 ,
the WRMS can be approximated by
#
$ N
$1
WRMS = % ∑ (wr s2r ) (2)
N r =1
where N is the number of samples in the signal used for RMS calculation and wr is the weighting factor
defined based on the number of consecutive sample r
wr = r m , m ≥ 0, r = 1, 2, . . . , N (3)
where m is a power of the weighting factor. The weighting factor with parameter m > 0 enables the
decrease of importance of the samples at the moment of excitation. The interpretation of WRMS maps
allows for the determination of the shape and location of the existing defects. In general, the areas with
higher WRMS values can indicate the possible locations of damage that are related to the reduction of
material stiffness and may be in the form of cracks, corrosion, or delamination [26].
3.1. Dispersion Curves

Before analyzing the defects in the lap joints, the characteristics of Lamb waves were determined
for two specimens: a single steel plate with a thickness of 3 mm and dimensions of 240 mm × 300 mm,
and two steel plates that were adhesively bonded with an adhesive layer with a thickness of 0.2 mm.
In each plate, a five-cycle Hanning windowed sine function with center frequencies of 50 kHz, 100 kHz,
and 200 kHz was used as an excitation signal. Out-of-plane vibration signals were measured in
101 points that were evenly distributed along a line with a length of 10 cm. The wavenumber-frequency
relations (Figure 6) were obtained using 2D-FFT (cf. [23,38]). In the experimental results, theoretical
dispersion curves that were obtained using GUIGUW software [39] based on the semi-analytical
finite-element formulation [40] were superimposed.
In general, in the single plate, both S0 and A0 modes can propagate in the considered frequency
range up to 200 kHz. However, the applied actuator utilizing the thickness-wise expansion effect
enabled mainly exciting the fundamental A0 mode and the influence of the S0 mode was insignificant,
as it can be seen in Figure 6a. The group velocities of A0 mode were 2981.5 m/s, 2575.7 m/s, and
2062.7 m/s for the excitation frequencies of 200 kHz, 100 kHz, and 50 kHz, respectively, giving the
wavelengths values of 14.9 mm, 25.7 mm, and 41.3 mm. Figure 6b shows the dispersion curves for
two adhesively bonded plates. In this case, three modes could propagate in the frequency range up to
200 kHz (S0 , A0 and A1 ), however, in the performed experiment, the A0 mode was dominant. It was
noted that the A0 mode dispersion curves were similar for both plates, indicating that this mode will
transmit across the waveguide transitions. The group velocities of the A0 mode propagating in two
adhesively bonded plates were 2865.6 m/s, 2448.5 m/s, and 2045.8 m/s for the excitation frequencies
of 200 kHz, 100 kHz, and 50 kHz, respectively, giving the wavelengths values of 14.3 mm, 24.5 mm,
and 40.9 mm.
240
Appl. Sci. 2018, 8, 522
Figure 6. Experimental and theoretical dispersion curves obtained for different excitation frequencies
of 50 kHz, 100 kHz, and 200 kHz: (a) steel 3 mm plate (E = 195.2 GPa, v = 0.3, ρ = 7741.7 kg/m3 );
and (b) two adhesively jointed steel 3 mm plates (E = 195.2 GPa, v = 0.3, ρ = 7741.7 kg/m3 ) adhesively
bonded with the adhesive layer of a thickness of 0.2 mm (E = 5 GPa, v = 0.35, ρ = 1130 kg/m3 ).
3.2. SLDV Maps

Figure 7 shows snapshots of the guided wave propagation signals that were obtained in the
experimental scanning of the analyzed specimens. The SLDV maps are presented at the time instance
of 27.7 μs. Disturbances of the wavefront shape on the disbond area were visible for each specimen
from #2–#9 when compared with the intact specimen #1. The differences were more or less considerable,
depending on the geometry of the defect. The most significant disturbances were observed for the
specimens with stripped defects skewed and parallel to the primary direction of propagation (#4–#6).
For specimens that had the disbond area in the shape of a strip perpendicular to the direction of
propagation (#2, #7), the changes were not clearly visible. The exception was specimen #3, which had
no adhesive near the edge of the overlap and resulted in the limited appearance of the wavefront on the
overlap surface. Internal defects (#8, #9) were moderately detectable. In summary, the presence of the
defects in the damaged specimens was visible in the maps, but their geometry was not determinable.
Figure 7. Scanning laser Doppler vibrometry (SLDV) maps of considered joints at the time instance
27.7 μs (dimensions of the joint are given in meters; the adhesive was 0.06–0.12 m).
241
Appl. Sci. 2018, 8, 522
3.3. WRMS Imaging
3.3.1. Influence of Time Window and Weighting Factor

Figure 8 shows the WRMS that was calculated for specimen #5 using Equation (2). Different
values of the weighting factor and time window were applied. In all calculations, the beginning of the
time window t1 was always set as zero, so the length of the time window was T = t2 . The obtained
damage maps are presented as logarithmic values of WRMS. The analysis of the results led to the
conclusion that the time window is a significant factor for the readability of visualizations. According
to the visual assessment, the clearness of the WRMS maps increased with the increase in the time
window. For this reason, the longest time window (T = 3.0 ms) was tentatively chosen for the damage
imaging of the other specimens (as presented further). It is also worth noting that the weighting factor
was not negligible. The constant weighting factor (wr = 1) was not efficient when compared with
others that increased with time. The maps were not clear enough and the influence of the incident
wave at the excitation point was highly visible. These effects were not observable for the weighting
factors wr = r and wr = r2 , as represented by the linear or square function of time, respectively. Their
effectiveness was comparable and significantly higher than the constant weighting factor.
Figure 8. Weighted root mean square (WRMS) maps (values in (m/s)) for specimen #5 for different
weighting factors wr = 1 (first column), wr = r (second column), wr = r2 (third column) and varying
time window: (a) T = 0.03 ms; (b) T = 0.05 ms; (c) T = 0.50 ms; (d) T = 1.00 ms; (e) T = 2.00 ms; and,
(f) T = 3.00 ms.
242
Appl. Sci. 2018, 8, 522
To assess the clearest variant of the WRMS, damage maps for adhesive joints with internal defects
(#8, #9) were prepared (Figure 9). The identification of the defect geometry for these specimens was
possible, however, some differences between the WRMS maps can be observed. A comparison of
the different weighting factors for both of the specimens showed the advantage of a variant with the
linear function.
Figure 9. WRMS maps (values in (m/s)) for joints with internal defects for constant time window
(T = 3.00 ms) and different weighting factors wr = 1 (first column), wr = r (second column), wr = r2
(third column): (a) specimen #8; and, (b) specimen #9.
3.3.2. Influence of Defect Shape and Position

Figure 10 shows the WRMS maps for all of the considered specimens #1–#9. The parameters of the
WRMS calculations were assumed as follows: the time window T = 3.00 ms and the weighting factor
(wr = r). The intensification of the WRMS maps could be observed at the disbond areas, where the
values were greater than for the properly bonded parts. This means that the amplitudes of propagating
waves vary more significantly if there is no direct connection between the bonded parts. It is worth
noting that the clearness of the maps increased with the difference between values that were obtained
for the disbonded and properly bonded areas.
Figure 10. WRMS maps (values in (m/s)) for specimens #1–#9 (T = 3.00 ms, wr = r).
243
Appl. Sci. 2018, 8, 522
The presence of defects was possible to determine in all cases, moreover, their shapes were also
identifiable. The clearest damage image was obtained for specimens #4–#6, as well as specimens #2
and #3. Internal damages in specimens #7–#9 were also possible to identify; however, the damage
maps had poorer quality. It is worth noting that the geometry of the identified defects was not in
ideal agreement with the assumed one. In some cases, minor irregularities were visible, especially
for samples that were intended as symmetric. This was an effect of inaccuracies in the process of the
specimen preparation. The verification of the actual state of the adhesive layer was obtained after
separating the joints (cf. Figure 3). It is clear that the adhesive leakages to the areas that are covered
with PTFE tape were not identified in the maps, which means that the tape prevented the plates from
being connected. On the other hand, the kissing defects were clearly noticeable (cf. Figure 3).
Figure 11 also shows the WRMS maps for all of the considered specimens (#1–#9), but for the
square weighting factor (wr = r2 ). The visual assessment of the presented maps led to conclusions
that were similar to the ones resulting from analysis of maps for the linear weighting factor (wr = r).
The value of m power affected the clearness of images: the ones for the linear weighting factor appeared
to be clearer. Based only on the visual assessment of the defect images, it was not possible to judge
which variant of weighting factor provided the clearest results.
Figure 11. WRMS maps (values in (m/s)) for specimens #1–#9 (T = 3.00 ms, wr = r2 ).
3.3.3. Influence of Excitation Frequency

An additional aspect of the analysis was the influence of the excitation frequency on the obtained
results. Figures 12 and 13 show the WRMS maps for specimens #6 and #7 for the linear and square
weighting factors, respectively. The measurement of the joints was repeated three times, with different
excitation frequencies, namely 200 kHz, 100 kHz, and 50 kHz. It was observed that the excitation
frequency was an important factor that significantly affected the readability of the maps. The efficiency
of the WRMS damage imaging increased with the increase in the excitation frequency value.
244
Appl. Sci. 2018, 8, 522
Figure 12. WRMS maps (values in (m/s)) with linear weighting factor (wr = r), time window T = 3.00 ms
and with different excitation frequencies f = 200 kHz (first column), f = 100 kHz (second column),
f = 50 kHz (third column): (a) specimen #6; and (b) specimen #7.
Figure 13. WRMS maps (values in (m/s)) with square weighting factor (wr = r2 ), time window
T = 3.00 ms and with different excitation frequencies f = 200 kHz (first column), f = 100 kHz
(second column), f = 50 kHz (third column): (a) specimen #6; and, (b) specimen #7.
3.3.4. WRMS Histograms

The WRMS method enables one to differentiate the areas of a properly prepared adhesive film and
its defects. In addition to the visual assessment of prepared maps, a numerical analysis of its calculated
values can be performed with the use of statistical analysis. The purpose is to prepare a histogram of
calculated WRMS values that is expected to reveal the bimodal distribution of the sample population
(concentrated on two values responding to disbonded and properly bonded areas). This presents the
possibility for the quantitative determination of the optimal values of WRMS parameters. As observed
from the images in Figures 10 and 11, higher WRMS values were obtained in the disbonded areas,
whereas the lower ones occurred in the areas of good adhesion. To provide a clear interpretation of the
WRMS histograms, the analysis was limited to the surface of the overlap (area of the adhesive joint).
In the first step, the influence of the time window on WRMS histograms was considered.
Figures 14 and 15 present the WRMS histograms for specimen #5 with different values of time
window and different weighting factors wr = r and wr = r2 , respectively. Additionally, to emphasize
the characterization of the obtained distributions, each histogram was approximated by the probability
density function with the use of the kernel smoother in a MATLAB® environment (shown as a red line).
It was clear that, for higher time window values (0.50 ms and more), a population of WRMS values
formed a bimodal distribution that was represented by a function with two conspicuous peaks values.
As concluded before, the clearness of the defect image was related to the difference between the values
that were obtained for health and poor adhesion areas of the overlap. For the histograms, this difference
was interpreted as the distance between the peak values that were measured along the horizontal axis.
For each diagram (if possible), peak values were localized. To compare the results that were obtained
245
Appl. Sci. 2018, 8, 522
for different time window values, the differences between the scales were eliminated by dividing the
measured distances by the analyzed range of fitted distributions. Calculated values for both variants
of weighting factors are presented in Table 2 (only if the bimodal distribution was obtained). It could
be seen that the increase in time window caused the peaks to occur at longer distances. The clarity of
the maps increased with the growth of the distance between peaks, which meant that the longer time
window positively influenced the clearness of the maps. This fact justifies the selection of the largest
value of time window (T = 3.00 ms) for the prepared WRMS maps.
Figure 14. Histograms of WRMS values (wr = r) for specimen #5 with different time windows:
(a) T = 0.03 ms; (b) T = 0.05 ms; (c) T = 0.50 ms; (d) T = 1.00 ms; (e) T = 2.00 ms; and, (f) T = 3.00 ms.
Figure 15. Histograms of WRMS values (wr = r2 ) for specimen #5 with different time windows:
(a) T= 0.03 ms; (b) T = 0.05 ms; (c) T = 0.50 ms; (d) T = 1.00 ms; (e) T = 2.00 ms; and, (f) T = 3.00 ms.
246
Appl. Sci. 2018, 8, 522
Table 2. Calculated values of relative peaks distance of bimodal distribution for specimen #5.
Time Window (ms) Relative Peaks Distance (wr = r) (%) Relative Peaks Distance (wr = r2 ) (%)
0.03 – –
0.05 – –
0.50 18.67 12.67
1.00 30.67 33.33
2.00 34.67 34.67
3.00 35.33 37.33
To move forward, histograms for all of the specimens were prepared with the time window
T = 3.00 ms. Figure 16 presents the histograms with the linear weighting factor. For the intact joint
(specimen #1), a common unimodal distribution was obtained. The calculated values were concentrated
about a singular mode that represented the undamaged area of the adhesive joint. Typical bimodal
distributions were obtained for specimens with a clear defect image (#2–#6). For specimens #7–#9,
the distributions were likely to be unimodal, but when compared with specimen #1, their shapes were
not that regular, and it was clear that their widths were greater than that of specimen #1. This may
result in the conclusion that the defect was present, but it could not be clearly observable. Figure 17
shows the histograms with the square weighting factor. The conclusions from the analysis of each
histogram were similar to ones based on Figure 16.
Figure 16. Histogram of WRMS values for specimens #1–#9 (T = 3.00 ms, wr = r).
247
Appl. Sci. 2018, 8, 522
Figure 17. Histogram of WRMS values for specimens #1–#9 (T = 3.00 ms, wr = r2 ).
Figure 18 shows the relationships between the relative peak distance and the weighting factor
that is represented by the power m for specimens #2–#6 that clearly presented bimodal distributions.
It could be seen that for most of the specimens, higher values of power m resulted in the greater value
of relative peaks distance. This means that the higher the value of m, the clearer the image of the defect
on the WRMS maps. It is worth noting that the value of the relative peaks distance stabilized at a
specific value of m individually for each defect, so the increase in m could not infinitely the improve
the quality of maps. Furthermore, specimen #4 provided a different relationship than the others where
the relative distance between the peaks at the beginning increased with power m and then decreased.
Maximal values were observed for the power m between 1 and 1.5. Based on the calculations presented
above, it was concluded that the proposed analysis provides the opportunity to improve the process
of determining the optimal value of the weighting factor (in fact, the power m) for damage imaging
using WRMS.
An attempt to quantitatively analyze the defects was performed based on the statistical analysis
of the calculated WRMS values. Histograms and their approximations by probability density functions
were prepared for different values of power m for specimens with a bimodal distribution (#2–#6).
The limit point (here, the central point) in the peaks distance was determined as the mean value
between peak values (along the horizontal axis). As previously mentioned, poor adhesion areas are
related to higher values of WRMS. For that reason, the number of values that are above the central point
value were counted, representing the number of points where the defect occurred. The percentage
of points with higher values can be interpreted as the relative surface of the defect. Based on the
relationships that are presented in Figure 19, it was clear that for most of the specimens, the relative
defect surface did not vary significantly in relation to power m and it moderately decreased with
its value. The exception was the defect in specimen #4 that presented the opposite relationship.
248
Appl. Sci. 2018, 8, 522
When comparing the results with measurements that were conducted on the real separated specimen
(Table 1), it was clear that the good agreement of the proposed approach was obtained only for defect
#3 (calculated mean value of relative surface is 68.5% and measured is 68.21%). For other specimens,
the differences were much higher (calculated values ranged from 45% to 55% and measured between
33% and 37%). The reason for these discrepancies may possibly be related to the method of determining
the limit point for counting the number of points that are related to the defect surface. Due to the
complexity of bimodal distributions (e.g., different parameters of both modes), further analyses should
be conducted.
Figure 18. Relationships between the relative peak distance of WRMS histograms and weighting factor
(represented by power m) for specimens #2–#6.
Figure 19. Relationships between the relative defect surface and weighting factor (represented by
power m) for specimens #2–#6.
One additional aspect of this research included the statistical analysis of the influence of the
excitation frequency. Figure 20 shows the WRMS histograms for specimen #6 with different excitation
frequencies. For the lowest value (50 kHz), the unimodal (but asymmetric) distribution was obtained.
Higher frequencies (100 kHz and 200 kHz) showed typical bimodal distributions with different
249
Appl. Sci. 2018, 8, 522
characterization. To compare them, the relative peak distance was calculated. For 100 kHz, a value
of 34.67% was obtained, whereas for 200 kHz, it was 27.33%. This meant that the 100 kHz frequency
provided results with more differentiated WRMS values obtained in areas with good and poor quality
(resulting in clearer defect image). Furthermore, the percentage of the defect surface was determined
with the use of the above-presented method with the central limit point. For 100 kHz, the result was
35.67% and for 200 kHz, a value of 51.78% was achieved. When compared to the measured value
(33.33%, Table 1), the nearest value was obtained for 100 kHz. However, due to the lack of results for
other specimens revealing bimodal distribution, it was not possible to conclude on the supremacy of
this frequency. Further research could help to solve this problem.
Figure 20. Histograms of WRMS values for specimen #6 with different excitation frequencies (T = 3.00 ms,
wr = r).
4. Conclusions
In this paper, an experimental analysis of the Lamb wave interaction with damage in the
adhesively bonded joints was conducted. The study consisted of a visual assessment and statistical
analysis of the obtained results. The presented investigations support the following conclusions:
• The analysis of maps that were obtained from the SLDV measurements provided the initial
identification of defects with different efficiencies. The clearness of the SLDV maps varied
depending on the shape and position of damage.
• The actual defect shape and location were determinable based on the appropriate damage imaging,
where recorded out-of-plane vibrations were processed by means of the WMRS. The effectiveness
of the applied method was strongly dependent on the calculation parameters (weighting factor
and time window), excitation frequency, and a type of damage.
• The extension of the time window led to the increase in differences in the WRMS values between
the damaged and undamaged areas of the adhesive joint. As a result, defects in the adhesive film
could be identified more accurately.
• By comparing different variants of weighting factors, the constant one did not provide efficient
results due to the high influence of the incident wave at the excitation point. This effect was
reduced by the weighting factor, which was represented by the linear or square function that
achieved comparable results. For specimens with internal defects, moderately greater effectiveness
of the linear variant was observed.
• The shape and location of defects had a great impact on the effectiveness of their identification.
Both the presence and the size of defects were possible to determine in all of the contested
specimens based on the WRMS damage maps. The clearest results were obtained for stripe defects
parallel or skewed to the primary direction of wave propagation. The disbonded areas in the
shape of the strip perpendicular to the direction of wave propagation were also determinable,
however, the identification was easier when such area was at the beginning or at the end of
the lap. The most difficult to detect were the defects in the internal part of the adhesive film.
250
Appl. Sci. 2018, 8, 522
The differentiation of effectiveness of the method for different types of defects resulted from the
occurrence of different wave phenomena at the boundaries of defects. Stripe defects, skewed,
or parallel to the direction of propagation triggered a diffraction of wave fronts that provided a
significant disturbance in wave propagation and resulted in a higher intensification of signals.
Defects in the shape of a perpendicular strip were related to wave reflection that did not disrupt
the propagation as diffraction did. Finally, internal defects were the most difficult to identify
because of the local characterization of occurring wave phenomena.
• The statistical analysis of the calculated WRMS values successfully supplemented the visual
assessment of the damage images. The interpretation of the histograms helped to explain the
effects that were observed on the WRMS maps. The use of the relative distance between peaks
of bimodal distribution was proposed as the indicator for the optimal determination of WRMS
parameters. Based on this indicator, it was proven that the largest value of the time window
obtained damage maps with the highest clearness. Moreover, it was possible to determine the
range of the optimal value of the weighting factor and to quantify the defect areas. The statistical
analysis of the influence of the excitation frequency also enabled a comparison of the effectiveness
of the three different frequencies. Two higher frequencies providing bimodal distribution were
proven to be more effective than the lowest excitation frequency providing unimodal distribution.
The main conclusion was that the scanning laser Doppler vibrometry and the weighted root
mean square processing of guided wave signals could be successfully applied for damage imaging in
adhesive films. Additionally, the statistical analysis can be useful in the interpretation of the obtained
results. However, the conducted research is an initial step for further analyses that are related to
condition assessment in adhesive joints of elements in real engineering structures. The proposed
method to determine the optimal value of the weighting factor and to quantify the real defect areas
could be further developed. The influence of different excitation frequencies could be also considered
more precisely.
Acknowledgments: The research work was carried out within project No. 2015/19/B/ST8/00779, financed by
the National Science Centre, Poland. This support is greatly acknowledged by the authors of the study.
Author Contributions: M.R. conceived, designed, and performed the experiments; J.L. contributed analysis tools
for damage imaging; M.R., E.W. and J.L. analyzed the data; E.W. and M.R. wrote the paper.
References
1. Adams, R.D.; Wake, W.C. Structural Adhesive Joints in Engineering; Elsevier Applied Science Publishers:
London, UK, 1986; ISBN 978-94-010-8977-7.
2. Karachalios, E.F.; Adams, R.D.; Da Silva, L.F.M. Strength of single lap joints with artificial defects. Int. J.
Adhes. Adhes. 2013, 45, 69–76. [CrossRef]
3. Heidarpour, F.; Farahani, M.; Ghabezi, P. Experimental investigation of the effects of adhesive defects on the
single lap joint strength. Int. J. Adhes. Adhes. 2018, 80, 128–132. [CrossRef]
4. Jeenjitkaew, C.; Guild, F.J. The analysis of kissing bonds in adhesive joints. Int. J. Adhes. Adhes. 2017, 75,
5. Adams, R.D.; Drinkwater, B.W. Nondestructive testing of adhesively-bonded joints. NDT E Int. 1997, 30,
93–98. [CrossRef]
6. Korzeniowski, M.; Piwowarczyk, T.; Maev, R.G. Application of ultrasonic method for quality evaluation of
adhesive layers. Arch. Civ. Mech. Eng. 2014, 14, 661–670. [CrossRef]
7. Vijayakumar, R.L.; Bhat, M.R.; Murthy, C.R.L. Non destructive evaluation of adhesively bonded carbon fiber
reinforced composite lap joints with varied bond quality. AIP Conf. Proc. 2012, 1430, 1276–1283. [CrossRef]
8. Tighe, R.C.; Dulieu-Barton, J.M.; Quinn, S. Identification of kissing defects in adhesive bonds using infrared
thermography. Int. J. Adhes. Adhes. 2016, 64, 168–178. [CrossRef]
9. Vijaya Kumar, R.L.; Bhat, M.R.; Murthy, C.R.L. Evaluation of kissing bond in composite adhesive lap joints
using digital image correlation: Preliminary studies. Int. J. Adhes. Adhes. 2013, 42, 60–68. [CrossRef]
251
Appl. Sci. 2018, 8, 522
10. Vijaya kumar, R.L.; Bhat, M.R.; Murthy, C.R.L. Analysis of composite single lap joints using numerical and
experimental approach. J. Adhes. Sci. Technol. 2014, 28, 893–914. [CrossRef]
11. Roth, W.; Giurgiutiu, V. Structural health monitoring of an adhesive disbond through electromechanical
impedance spectroscopy. Int. J. Adhes. Adhes. 2017, 73, 109–117. [CrossRef]
12. Giurgiutiu, V. Structural Health Monitoring with Piezoelectric Wafer Active Sensors; Elsevier: Amsterdam,
The Netherlands, 2008.
13. Ostachowicz, W.; Kudela, P.; Krawczuk, M.; Zak, A. Guided Waves in Structures for SHM; Wiley: Hoboken, NJ,
USA, 2012; ISBN 9781119965855.
14. Rose, J.L. Ultrasonic Guided Waves in Solid Media; Cambridge University Press: New York, NY, USA, 2014;
ISBN 9781107273610.
15. Rokhlin, S. Lamb wave interaction with lap-shear adhesive joints: Theory and experiment. J. Acoust. Soc. Am.
1991, 89, 2758–2765. [CrossRef]
16. Rose, J.L.; Rajana, K.M.; Hansch, M.K.T. Ultrasonic Guided Waves for NDE of Adhesively Bonded Structures.
J. Adhes. 1995, 50, 71–82. [CrossRef]
17. Lowe, M.J.S.; Challis, R.E.; Chan, C.W. The transmission of Lamb waves across adhesively bonded lap joints.
J. Acoust. Soc. Am. 2000, 107, 1333–1345. [CrossRef] [PubMed]
18. Di Scalea, F.L.; Bonomo, M.; Tuzzeo, D. Ultrasonic guided wave inspection of bonded lap joints: Noncontact
method and photoelastic visualization. Res. Nondestruct. Eval. 2001, 13, 153–171. [CrossRef]
19. Ren, B.; Lissenden, C.J. Ultrasonic guided wave inspection of adhesive bonds between composite laminates.
Int. J. Adhes. Adhes. 2013, 45, 59–68. [CrossRef]
20. Sherafat, M.H.; Quaegebeur, N.; Lessard, L.; Masson, P.; Hubert, P. Guided Wave Propagation through
Composite Bonded Joints. In Proceedings of the EWSHM 7th European Workshop on Structural Health
Monitoring, Nantes, France, 10 July 2014.
21. Ren, B.; Lissenden, C.J. Modal content-based damage indicators for disbonds in adhesively bonded composite
structures. Struct. Health Monit. Int. J. 2016, 15, 491–504. [CrossRef]
22. Mori, N.; Biwa, S. Transmission of Lamb waves and resonance at an adhesive butt joint of plates. Ultrasonics
2016, 72, 80–88. [CrossRef] [PubMed]
23. Gauthier, C.; El-Kettani, M.; Galy, J.; Predoi, M.; Leduc, D.; Izbicki, J.L. Lamb waves characterization of
adhesion levels in aluminum/epoxy bi-layers with different cohesive and adhesive properties. Int. J.
Adhes. Adhes. 2017, 74, 15–20. [CrossRef]
24. Mallet, L.; Lee, B.C.; Staszewski, W.J.; Scarpa, F. Structural health monitoring using scanning laser vibrometry:
II. Lamb waves for damage detection. Smart Mater. Struct. 2004, 13, 261–269. [CrossRef]
25. Kudela, P.; Wandowski, T.; Malinowski, P.; Ostachowicz, W. Application of scanning laser Doppler
vibrometry for delamination detection in composite structures. Opt. Lasers Eng. 2016, 99, 46–57. [CrossRef]
26. Pieczonka, Ł.; Ambroziński, Ł.; Staszewski, W.J.; Barnoncel, D.; Pérès, P. Damage detection in composite
panels based on mode-converted Lamb waves sensed using 3D laser scanning vibrometer. Opt. Lasers Eng.
2017, 99, 80–87. [CrossRef]
27. Sunarsa, T.Y.; Aryan, P.; Jeon, I.; Park, B.; Liu, P.; Sohn, H. A reference-free and non-contact method for
detecting and imaging damage in adhesive-bonded structures using air-coupled ultrasonic transducers.
Materials (Basel) 2017, 10, 1402. [CrossRef]
28. Esfandabadi, Y.K.; De Marchi, L.; Testoni, N.; Marzani, A.; Masetti, G. Full wavefield analysis and Damage
imaging through compressive sensing in Lamb wave inspections. IEEE Trans. Ultrason. Ferroelectr.
Freq. Control 2017, 65, 269–280. [CrossRef] [PubMed]
29. Dziendzikowski, M.; Dragan, K.; Katunin, A. Localizing impact damage of composite structures with
modified RAPID algorithm and non-circular PZT arrays. Arch. Civ. Mech. Eng. 2017, 17, 178–187. [CrossRef]
30. Żak, A.; Radzieński, M.; Krawczuk, M.; Ostachowicz, W. Damage detection strategies based on propagation
of guided elastic waves. Smart Mater. Struct. 2012, 21, 35024. [CrossRef]
31. Radzieński, M.; Doliński, Ł.; Krawczuk, M.; Palacz, M. Damage localisation in a stiffened plate structure
using a propagating wave. Mech. Syst. Signal Process. 2013, 39, 388–395. [CrossRef]
32. Lee, C.; Park, S. Flaw Imaging Technique for Plate-Like Structures Using Scanning Laser Source Actuation.
Shock Vib. 2014, 725030. [CrossRef]
252
Appl. Sci. 2018, 8, 522
33. Jothi Saravanan, T.; Gopalakrishnan, N.; Prasad Rao, N. Damage detection in structural element through
propagating waves using radially weighted and factored RMS. Meas. J. Int. Meas. Confed. 2015, 73, 520–538.
[CrossRef]
34. Geetha, G.K.; Roy Mahapatra, D.; Gopalakrishnan, S.; Hanagud, S. Laser Doppler imaging of delamination in
a composite T-joint with remotely located ultrasonic actuators. Compos. Struct. 2016, 147, 197–210. [CrossRef]
35. Marks, R.; Clarke, A.; Featherston, C.; Paget, C.; Pullin, R. Lamb Wave Interaction with Adhesively Bonded
Stiffeners and Disbonds Using 3D Vibrometry. Appl. Sci. 2016, 6, 12. [CrossRef]
36. Lee, C.; Zhang, A.; Yu, B.; Park, S. Comparison study between RMS and edge detection image processing
algorithms for a pulsed laser UWPI (Ultrasonic wave propagation imaging)-based NDT technique.
Sensors (Switzerland) 2017, 17, 1224. [CrossRef] [PubMed]
37. Abràmoff, M.D.; Magalhães, P.J.; Ram, S.J. Image processing with imageJ. Biophotonics Int. 2004, 11, 36–41.
[CrossRef]
38. Alleyne, D.N.; Cawley, P. A 2-dimensional Fourier transform method for the quantitative measurement of
Lamb modes. In Proceedings of the IEEE Ultrasonics Symposium, Honolulu, HI, USA, 4–7 December 1990;
Volume 2, pp. 1143–1146. [CrossRef]
39. Bocchini, P.; Asce, M.; Marzani, A.; Viola, E. Graphical User Interface for Guided Acoustic Waves. J. Comput.
Civ. Eng. 2011, 25, 202–210. [CrossRef]
40. Mazzotti, M.; Bartoli, I.; Marzani, A.; Viola, E. A coupled SAFE-2.5D BEM approach for the dispersion
analysis of damped leaky guided waves in embedded waveguides of arbitrary cross-section. Ultrasonics
2013, 53, 1227–1241. [CrossRef] [PubMed]
253
applied
sciences
Article
Rapid High-Resolution Wavenumber Extraction from
Ultrasonic Guided Waves Using Adaptive Array
Signal Processing
Shigeaki Okumura 1, *, Vu-Hieu Nguyen 2 , Hirofumi Taki 3 , Guillaume Haïat 4 , Salah Naili 2
and Toru Sato 1
1 Graduate School of Informatics, Kyoto University, Kyoto 606-8501, Japan; sato.toru.6e@kyoto-u.ac.jp
2 Université Paris-Est, Laboratoire Modélisation et Simulation Multi Echelle, MSME UMR 8208 CNRS,
94010 Créteil CEDEX, France; vu-hieu.nguyen@u-pec.fr (V.-H.N.); naili@u-pec.fr (S.N.)
3 Graduate School of Biomedical Engineering, Tohoku University, Sendai 980-8575, Japan;
hirofumi.taki@mb6.seikyou.ne.jp
4 CNRS, Laboratoire Modélisation et Simulation Multi Echelle, MSME UMR 8208 CNRS,
94010 Créteil CEDEX, France; haiat@u-pec.fr
* Correspondence: sokumura@sato-lab.0t0.jp; Tel.: +81-75-753-3394
Received: 6 March 2018; Accepted: 16 April 2018; Published: 23 April 2018
Abstract: Quantitative ultrasound techniques for assessment of bone quality have been attracting
significant research attention. The axial transmission technique, which involves analysis of ultrasonic
guided waves propagating along cortical bone, has been proposed for assessment of cortical
bone quality. Because the frequency-dependent wavenumbers reflect the elastic parameters of
the medium, high-resolution estimation of the wavenumbers is required at each frequency with low
computational cost. We use an adaptive array signal processing method and propose a technique
that can be used to estimate the numbers of propagation modes that exist at each frequency without
the need for time-consuming calculations. An experimental study of 4-mm-thick copper and
bone-mimicking plates showed that the proposed method estimated the wavenumbers accurately
with estimation errors of less than 4% and a calculation time of less than 0.5 s when using a
laptop computer.
Keywords: signal processing; ultrasonic guided waves; axial transmission
1. Introduction
The first quantitative ultrasound (QUS) technique for use in bone assessment was proposed by
Langton et al. [1] in the 1980s. Since then, a number of different QUS techniques have been proposed
and developed [2–4]. QUS techniques were developed for early osteoporosis detection and fracture risk
evaluation screening applications. When compared with X-rays, QUS offers several advantages: it is a
noninvasive process and can be performed using portable and low-cost equipment. Techniques that
can evaluate the quality of cortical long bone structures have recently attracted significant research
attention [2,3,5–8]. In this study, we have developed cortical long bone quality evaluation methods
using the axial transmission (AT) technique [9–13].
A number of recent studies have used the AT technique, in which wide-band signals are emitted and
the ultrasonic guided wave that propagates along the cortical bone is then analyzed. The cortical bone
has previously been described as a transversely isotropic absorbing plate with finite thickness [14–18].
The ultrasonic guided wave that propagates in cortical bone consists of multiple propagation modes, and
the frequency-dependent wavenumbers of these modes represent the elastic properties of the medium.
Many techniques for frequency-wavenumber ( f –k) analysis have therefore been proposed and reported.

Appl. Sci. 2018, 8, 652
Tran et al. proposed the Radon transform method, which estimates the phase velocity (cp ) at
each frequency [12]. The phase velocity is given by cp = 2π f /k, and the basic premises of the
f –k and f –cp analysis methods are theoretically the same. The Radon transform method uses a
single transmitter with multiple receivers, estimates the phase velocity using an iterative process,
and produces high-resolution estimates. However, the computational cost of this method is not low
because it requires inversion of a large-scale matrix. Sasso et al. [19] proposed a singular value
decomposition (SVD)-based method and used a probe composed of multiple transmitters and emitters
to analyze the guided waves. Minonzio et al. extended this method and produced accurate depictions
of the phase velocities of the plate [10]. Use of SVD allows the method to detect the modes with
low intensity. The method was effective for estimation of the elastic properties of the medium [20].
However, the measurement resolution is determined by the aperture size, and it is necessary to
estimate the number of propagation modes in the received signal. Xu et al. proposed the sparse SVD
(S-SVD) method, which combines the Radon transform method proposed by Tran et al. with the SVD
method [21]. The resulting method acquires super-resolution estimates using a combination of SVD
and an iterative process. However, this method also requires estimation of the number of propagation
modes in the received signal and multiple calculations for inversion of a large matrix.
We recently developed an adaptive beamforming technique [22]. We used the estimation of signal
parameters via rotational invariance technique (ESPRIT) algorithm. This algorithm uses eigenvalue
decomposition to separate the desired signal from noise. The ESPRIT algorithm also requires estimation
of the number of propagation modes and the conventional method includes an iterative process with
eigenvalue decomposition [22]. Because the matrix inversion and eigenvalue decomposition processes
incur large computational costs, reduction of these computational costs would be useful for practical
applications. Therefore, in this study, we propose a method that estimates the number of propagation
modes with a low computational cost.
The number of propagation modes in the received signal, which is denoted by M in this study,
is usually estimated using a thresholding process. However, the eigenvalues that correspond to the
signal are not equivalent to the signal intensities. Therefore, it is not easy to determine M when using a
thresholding process that only uses the intensities of the eigenvalues or singular values. Therefore, in this
study, we propose a new algorithm to estimate M using information theoretic criteria. This estimation
procedure would be effective for both the SVD and ESPRIT methods. While many studies on estimation of
M have been reported [23–27], the resulting methods were not applied to the analysis of ultrasonic guided
waves when propagating along a transversely isotropic absorbing material such as cortical bone.
In the proposed method, we do not use an iterative process with high computational cost, such as
the eigenvalue decomposition or matrix inversion methods, but use a diagonal loading (DL) technique
that adds the diagonal loading matrix to the covariance matrix to determine M. We compare the
computational cost of the proposed method with that of a conventional method and demonstrate
the effectiveness of the proposed method via simple numerical simulations and experiments using a
copper plate and a bone-mimicking plate.
2.1. System Model

In this study, we considered the system model that is shown in Figure 1. We placed a linear
array probe on top of a free isotropic or transversely isotropic plate and analyzed the guided waves
that propagate along the plate. When we consider multiple propagation signals with multiple phase
velocities and ignore changes in amplitude, the received signals are given by
M(ω )
S n ( ω ) = S1 ( ω ) ∑ exp{− j(n − 1)k m (ω ) x } (1)
m =1
256
Appl. Sci. 2018, 8, 652
where Sn (ω ) is the received signal at the n-th receiver in the frequency domain with an angular
frequency ω and x is the receiving array pitch. k m is the wavenumber of the m-th propagation mode,
which is given by k m = ω/cm , where cm is the phase velocity of the m-th mode. Note that M is
dependent on the frequency.
Linear array probe
Transmitter Receivers
Thickness Specimen
Figure 1. System model used in this study. In this study, we used a 4-mm-thick copper (homogeneous
isotropic) plate and a 4-mm-thick bone mimicking (transversely isotropic) plate.
2.2. Wavenumber Estimation Using the ESPRIT Algorithm

To realize super-resolution estimates, we used the ESPRIT algorithm. The basis of this algorithm
is briefly explained here. When there is a single propagating wave, i.e., M = 1 in Equation (1), we can
then estimate the wavenumber by comparing the signals that are received at two receivers.
k1 (ω ) = {Sn (ω )Sn∗ +1 (ω )}/x (2)
where represents the angle of the complex signal.

As shown in Equation (2), the wavenumber is estimated directly without performing a peak
search process. However, when M > 1, we cannot estimate the wavenumber by simply comparing
the phases of the two received signals alone. The ESPRIT algorithm estimates the wavenumbers of
multiple modes via an eigenvalue decomposition of the covariance matrix to enable the separation of
multiple signals.
The covariance matrix represents a correlation between the signals at each receiver. To estimate the
covariance matrix, we use the sub-array averaging technique [28]. In the sub-array averaging technique,
multiple sub-arrays are situated into the full-size array. Here, we define the signal matrix as follows:
⎛ ⎞
S1,1 (ω ) ··· S N,1 (ω )
⎜ .. .. . ⎟
A(ω ) = ⎝ . . .. ⎠ (3)
S1,Nsub (ω ) · · · S N,Nsub (ω )
where Si,j is the received signal at the j-th receiver in the i-th sub-array, N is the number of sub-arrays
contained within the whole array, Nsub is the number of receivers included within a single sub-array,
i.e., the number of whole receivers is NR = N + Nsub − 1, and A is the signal matrix with size Nsub × N.
The covariance matrix, R, is then given by
R(ω ) = A ( ω ) AH ( ω ) (4)
N
= ∑ Ri ( ω ) (5)
i =1
N
= ∑ Si (ω )SHi (ω ) (6)
i =1
257
Appl. Sci. 2018, 8, 652
where Si (ω ) = [Si,1 (ω ), · · · , Si,Nsub (ω )]T is the signal vector at the i-th sub-array and T and H denote
transpose and hermitian transpose of a matrix, respectively. As shown in Equations (5) and (6),
the covariance matrix can be estimated by averaging the covariance matrices of the sub-arrays.
Note that it is necessary to estimate M to enable estimation of the wavenumber.
The theoretical characteristics of the eigenvalues are expressed as follows:
l1 (ω ) ≥ · · · ≥ l M(ω ) (ω ) ≥ l M(ω )+1 (ω ) = · · · = l Nsub (ω ) = σn2 (7)
where li (ω ) is i-th eigenvalue and σn2 is the noise intensity. Note that the absolute values of these
eigenvalues do not match the signal intensity directly, i.e., when two waves have the same intensity,
the eigenvalues that correspond to these signals do not have the same value. Therefore, the simple
thresholding process is not suitable for accurate estimation of the number of signals.
In this section, we calculate the covariance matrix and apply the eigenvalue decomposition technique.
When we apply SVD to the signal matrix, A, we theoretically obtain the same result as that obtained
when using N transmitters and Nsub receivers with the SVD technique.
2.3. Estimation of the Number of Signals
2.3.1. Overview of the Basic Theory

In this section, we propose an estimation technique for M(ω ). A method to estimate M(ω )
that uses information theoretic criteria called the minimum description length (MDL) principle has
previously been reported [23,24]. The evaluation index G (m) that is used to estimate M is given by
⎛ ⎞( Nsub −m) N
1
⎜ Nsub ⎟
⎜ ∏i=m+1 li (ω ) Nsub − m ⎟ 1
G (m) = − log ⎜
⎜ N li ( ω )
⎟
⎟ + m(2Nsub − m) log N (8)
⎝ ∑i=sub ⎠ 2
m+1 Nsub −m
where G (m) is the index, and the value of m that minimizes G (m) represents the estimated M(ω ).
Here we define the estimated M(ω ) as M (ω ).
2.3.2. Diagonal Loading Technique

In Equation (8), the numerator and the denominator represent the geometric and arithmetic
means of the eigenvalues, respectively. Smaller eigenvalues therefore make the estimates unstable.
To stabilize the estimation procedure, we thus use a diagonal loading technique that adds a diagonal
matrix to the covariance matrix [25–27].
The DL process is expressed as follows:
R (ω ) = R(ω ) + η (ω )I (9)
where η is a diagonal loading factor and I is the identity matrix. The process that was shown in the
previous equation can then be rewritten as follows:
li (ω ) = li (ω ) + η (ω ) (10)
where li (ω ) is the i-th eigenvalue that is calculated from R (ω ). Therefore, even when a different
DL factor is used with the same covariance matrix, we do not need to perform the eigenvalue
decomposition process again. In other words, we use the DL technique to estimate M as an offset
value and do not add the DL matrix to the covariance matrix directly. By replacing li in Equation (8)
with li , we can then obtain the modified estimates.
258
Appl. Sci. 2018, 8, 652
2.3.3. Determination of the DL Factor

Selection of the DL factor is an important aspect of the estimation of M. In this study, we used
two DL factors that were dependent on the received signal intensity. A schematic illustration of these
two factors is shown in Figure 2.
Signal Intensity
Intensity
Proposed DL
Constant DL
Frequency dependent DL
Frequency
Figure 2. Proposed diagonal loading (DL) technique.
The DL factor is given by

η (ω ) = η1 (ω )δ1 (ω ) + η2 δ2 (11)
δ1 (ω ) = tr{R(ω )} (12)
' ω2
1
δ2 = tr{R(ω )}dω (13)
ω2 − ω1 ω1
where η1 and η2 are the two component DL factors. η2 δ2 is used for stabilization and is called
the constant DL in Figure 2 because, at low signal intensities, η1 (ω )δ1 (ω ), which is called the
frequency-dependent DL in Figure 2, approaches zero, and the estimates would thus be unstable.
To determine η1 (ω ) for the wavenumber estimation procedure, we assumed that the optimal
η1 (ω ) is common within a specific frequency range. This assumption was made because, as mentioned
below, M shows a step-like change with respect to η1 . The optimal η1 that gives the optimum M value
has a range and is not a critical value.
We vary η1 (ω ) and select the minimum value of η1 that gives M (ω ) such that it satisfies the
following condition:
max{ M (ω )} < Mth , with (ω − ωw ≤ ω ≤ ω + ωw ) (14)
where ωw is the width of the frequency window that is used for stable estimation and Mth is a threshold
M value that is sufficiently large for wavenumber estimation.

The setup of the experimental study is the same as the system model shown at Figure 1, but we used
a 128-element linear array probe with an element pitch of 0.375 mm that was manufactured by Japan Probe
(Kanagawa, Japan). We selected a single transmitter and 28 receivers (NR = 28) with an element pitch of
0.75 mm. The distance between the transmitter and the first receiver was 18.75 mm. The center frequency
of the transmitted wave was 1.0 MHz. We used two test specimens: (1) a 4-mm-thick copper plate (with
a shear wave velocity of 2260 m/s and a longitudinal wave velocity of 4650 m/s) and (2) a 4-mm-thick
transversely isotropic bone-mimicking plate (Sawbones, Vashon, WA, USA). The elastic parameters of
the bone-mimicking plate have been determined in previous studies [20,21]. The density, shear wave
velocity, longitudinal wave velocity along fibers, and longitudinal wave velocity orthogonal to the
fibers are 1.64 g·cm−3 , 1620 m/s, 3570 m/s, and 2910 m/s, respectively [20,21]. We put the probe
parallel to the fibers.
259
Appl. Sci. 2018, 8, 652
In the proposed method, we set Mth = Nsub − 1, ωw = 0.5 MHz, η2 = −40 dB, and Nsub = 15.
We used ω1 = 4.9 MHz and ω2 = 5.5 MHz to select a frequency range that should include only noise.
We prepared the DL factor using values of η1 within the −100 dB to 0 dB range at intervals of
10 dB. Mth = Nsub − 1 represents the maximum value of M when using a sub-array size of Nsub .
In the conventional SVD method, we used values of N = 5 and Nsub = 24 to match those used
in the conventional study [21]. In the previous study, multiple transmitters were used. The use of
multiple transmitters and of sub-arrays is theoretically the same.
3.1. Evaluation of the Number of the Propagation Modes with DL

To evaluate the M estimation procedure using the DL technique, we performed a simple
simulation because it is difficult to know ground truth of M in experimental study. We assumed
that three waves with k = 1000, 2000, and 3000 rad/m were propagating. The signal-to-noise ratio
(SNR) was 40 dB, NR = 28, and Nsub = 15.
The estimation results showed step-like changes with increasing η1 . Within the
−100 dB ≤ η1 ≤ −90 dB, −80 dB ≤ η1 ≤ −20 dB, and −10 dB ≤ η1 ≤ 0 dB ranges, the corresponding
estimated M values were 14, 3, and 0, respectively. The root mean square error (RMSE) of the
wavenumber estimation process with η1 that gives the correct M is 0.73 rad/m. An important point
that should be noted here is that the true M value is estimated using a wide range for η1 , e.g.,
−80 dB ≤ η1 ≤ −20 dB.

We first investigate the optimal size for the sub-array. Figure 3 shows the RMSEs when we use
full-array sizes of NR = 28 and 32 and change sub-array size. The measured specimen was bone-mimicking
plate. The results show that the optimum sub-array size is N = NR /2. Therefore, with NR = 28, the value
of Nsub = 15 is the optimal size.
350
300
RMSE (rad/m)
250
200
N R =28
150 N R =32
N= N R /2
100
5 10 15 20 25
Nsub
Figure 3. Dependence of the proposed method on sub-array size. The black solid line and the dotted
line show the results for arrays composed of different numbers of receivers. The vertical black dotted
lines show the horizontal values with N = NR /2, i.e., Nsub − 1 = NR /2.
In the previous paragraph, we described the experimental investigation of the optimal size.
Here, we describe the theoretical investigation of this size [29]. To recover the rank of the covariance
matrix with M modes, a minimum of M times averaging, i.e., N = M, is required. In addition,
to measure M modes, a minimum of Nsub = M + 1 is required. Under these two conditions,
when we attempt to maximize M and Nsub , the number of receivers NR should be expressed using
NR = Nsub + N − 1 = 2N. Theoretically, we can confirm that the selection N = NR /2 is reasonable.
260
Appl. Sci. 2018, 8, 652
We determined that Nsub = 15 is the optimal size. Therefore, the following results employed
Nsub of 15. Figures 4–7 show the experimental results that were obtained for the copper plate and the
bone-mimicking plate, respectively. The color maps show the results obtained using the conventional
SVD method. The red dots represent estimated results obtained using the proposed method. The solid
white lines show the theoretical curves. To remove any obvious false estimates, we removed all estimates
with cp < 1200 and cp > 50, 000 m/s. The proposed method thus successfully depicted the wavenumber
without the need for peak search processes.
The RMSEs of the estimation results when using the proposed method with the copper plate
and bone-mimicking plate were 108 rad/m and 121 rad/m, respectively. RMSEs are shown in
Table 1. The wavenumbers of the shear waves that had larger effects on the theoretical curve than the
longitudinal wave at the center frequency (1.0 MHz) of the copper plate and the bone-mimicking plate
were 2780 rad/m and 3879 rad/m, respectively. When compared with the wavenumber of the shear
wave, the RMSE was less than 4%.
6000 1
5000
0.8
Wavenumber(rad/m)
4000
0.6
3000
0.4
2000
0.2
1000
0 0
0.5 1 1.5 2
Frequency(MHz)
Figure 4. Wavenumbers of the 4-mm-thick copper plate estimated using the conventional singular
value decomposition (SVD) method. The color intensity map shows the results obtained using the
conventional SVD method. Solid white lines show the theoretical curves.
6000 1
5000
0.8
Wavenumber(rad/m)
4000
0.6
3000
0.4
2000
0.2
1000
0 0
0.5 1 1.5 2
Frequency(MHz)
Figure 5. Wavenumbers of the 4-mm-thick copper plate estimated using the proposed method and
the conventional SVD method. Red dots show the estimates obtained using the proposed method.
The color intensity map shows the results obtained using the conventional SVD method. Solid white
lines show the theoretical curves.
261
Appl. Sci. 2018, 8, 652
6000 1
5000
0.8
Wavenumber(rad/m)
4000
0.6
3000
0.4
2000
0.2
1000
0 0
0.5 1 1.5 2
Frequency(MHz)
Figure 6. Wavenumbers of the 4-mm-thick bone-mimicking plate estimated using the conventional
SVD method. The color intensity map shows the results obtained using the conventional SVD method.
Solid white lines show the theoretical curves.
6000 1
5000
0.8
Wavenumber(rad/m)
4000
0.6
3000
0.4
2000
0.2
1000
0 0
0.5 1 1.5 2
Frequency(MHz)
Figure 7. Wavenumbers of the 4-mm-thick bone-mimicking plate estimated using the proposed method
and the conventional SVD method. Red dots show the estimates obtained using the proposed method.
The color intensity map shows the results obtained using the conventional SVD method. Solid white
lines show the theoretical curves.
Table 1. Root mean square errors (RMSEs) of the experimental results with different settings. The units
are rad/m.
Proposed Method with Proposed Method with ESPRIT with Fixed ESPRIT with Fixed Proposed Method with Proposed Method with
Copper Plate Bone-Mimicking Plate Threshold (−40 dB) Threshold (−30 dB) ωw = 0.1 MHz ωw = 0.75 MHz
108 121 184 143 144 118
The spectrum at the frequency that is indicated by the white dotted line in Figure 7 is shown in
Figure 8. As Figure 8 illustrates, the proposed method has higher resolution than the conventional method
because the method can depict two modes around 4000 rad/m with small error. The wavenumbers that
were estimated using the proposed method were almost the same as those at the peak position in the
SVD spectrum when the resolution of the SVD method was high enough or the resolution of the ESPRIT
algorithm was insufficient to separate the signals. However, it is difficult to identify some of the peaks
because of their small amplitudes and low prominence when using the SVD method, such as the peaks
with wavenumbers of approximately 2800 and 4300 rad/m.
262
Appl. Sci. 2018, 8, 652
Spectrum of the conventional SVD

Proposed method
True wavenumber
Normalized amplitude
0.8
0.6
0.4
0.2
0
0 1000 2000 3000 4000 5000 6000
Wavenumber (rad/m)
Figure 8. Depiction of the wavenumbers obtained using the proposed method and the spectrum
obtained using the conventional method at 1.0 MHz. This is the cross-sectional view at the white dotted
line shown in Figure 7. The vertical blue dotted lines indicate the theoretical values.
The wavenumber or phase velocity estimation steps that have high computational
costs in the proposed method are eigenvalue decomposition of the covariance matrix with
a size of Nsub × Nsub , O( Nsub 3 ) and the SVD of the matrix of size N
sub − 1 × 2M,
* +
O(min ( Nsub − 1) × 4M2 , ( Nsub − 1)2 × 2M ). The calculation steps are shown in detail in [30].
The corresponding steps in the conventional SVD method, the Radon transform method, and the
S-SVD method are the SVD of the matrix of size N × Nsub , O( N 2 × Nsub ) [30], the inversion of the
matrix of size Nc × Nc , O( Nc3 ), and the inversion of the matrix of size Nk × Nk , O( Nk3 ), respectively,
where Nc and Nk are the numbers of sampling points in the phase velocity and wavenumber axes,
which is normally larger than 64. Therefore, the SVD method showed the smallest computational cost
with the lowest resolution. The computational cost of the proposed method was significantly lower
than the corresponding costs of the Radon and S-SVD methods.
The calculation time (not including loading time) for the experimental data was less than 0.5 s
when using a commercial processor (Core(TM) i7-7500U; Intel, Santa Clara, CA, USA) on a laptop
computer with MATLAB (R2017a, Mathworks, Natick, MA, USA) platform. The calculation time taken
for the conventional Radon transform was less than 60 s, according to [12]. While noting here that
we did not use the same processor or computer, and in fact used a faster processor, the reduction in
the calculation time of 99% is interesting. In addition, the computational cost of the S-SVD method
is greater than that of the Radon method. The calculation time of the conventional SVD method is
around 0.1 s. Note that, as mentioned above, the conventional SVD method has considerably lower
computational complexity than the proposed method. Because we ran the SVD method and the
proposed method on MATLAB platform, the calculation time does not directly reflect the theoretical
complexity. To examine the effectiveness of the proposed method in determination of M, we compared
the performance of the proposed method with that of the ESPRIT method with a fixed threshold.
Figures 9 and 10 show the results obtained for a bone plate when we used eigenvalues that exceed −40
and −30 dB of the maximum eigenvalue, respectively. The red circles represent the results obtained
using the proposed method. The blue cross marks show the estimates obtained using the ESPRIT
algorithm with a fixed threshold. The solid gray lines show the theoretical curves. The RMSEs of
the ESPRIT method with fixed thresholds of −40 and −30 dB were 184 and 143 rad/m, respectively.
As shown in Figure 9, smaller thresholds tend to make the estimates unstable because the ESPRIT
algorithm treated the noise as the desired signal. In contrast, larger thresholds can miss the weaker
modes as shown in Figure 10. The threshold should thus be determined manually. The results are
sensitive to the threshold value, so it is difficult to determine the threshold.
263
Appl. Sci. 2018, 8, 652
ESPRIT method with fixed threshold of -40 dB

Proposed method
6000
5000
Wavenumber(rad/m)
4000
3000
2000
1000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Frequency(MHz)
estimation of signal parameters via rotational invariance technique (ESPRIT) algorithm at a fixed
threshold of −40 dB and the proposed method. Red dots show the estimates obtained using the
proposed method and blue cross marks show the estimates obtained using the conventional method.
Solid gray lines show the theoretical curves.
ESPRIT method with fixed threshold of -30 dB

Proposed method
6000
5000
Wavenumber(rad/m)
4000
3000
2000
1000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Frequency(MHz)
ESPRIT algorithm at a fixed threshold of −30 dB and the proposed method. Red dots show the
estimates obtained using the proposed method and blue cross marks show the estimates obtained
using the conventional method. Solid gray lines show the theoretical curves.
Figure 11 shows the wavenumbers for values of Nsub = 5 and 26. The red cross marks and blue
circles show the results obtained for Nsub = 5 and 26, respectively. The black lines show the theoretical
curves. The smaller sub-array had lower resolution and missed many of the modes, while the larger
sub-array caused false estimates.
We used a frequency window, denoted by ωw , of 0.5 MHz. When we used ωw values of 0.1
and 0.75 MHz in the experiments with the bone-mimicking plate, the RMSEs obtained were 144 and
118 rad/m, respectively. Therefore, the dependence on ωw was insignificant and a larger window
width provided stable estimation.
We chose a DL factor of η2 = −40 dB. Because this value was added for stabilization when we
analyzed the frequency range using a significantly smaller signal intensity, a smaller value was selected.
When we used a larger value, more stable estimates were obtained as a result but the weaker modes
were missed by the method.
264
Appl. Sci. 2018, 8, 652
6000
N sub =5
5000 N sub =26
Wavenumber(rad/m)
4000
3000
2000
1000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Frequency(MHz)
Figure 11. Wavenumbers estimated for differently sized sub-arrays. Red cross marks and blue circles
show the results obtained for Nsub = 5 and 26, respectively. Black lines show the theoretical curves.
We used the ESPRIT algorithm in this study. The novel advantage of the ESPRIT algorithm when
used for wavenumber estimation of a guided wave propagating along the cortical bone is that it can
estimate the wavenumber without the need for a peak search process. To extract the wavenumber from
the f –k spectrum image when using conventional methods such as the Radon and SVD-based methods,
multiple one-dimensional peak search processes are required [6,20]. While the computational cost is not
high, parameters such as the intensity threshold and prominence must be set. Thus, the performance
of the conventional method such as RMSE depends on the additional parameter setting.
4. Conclusions
In this study, we proposed a high-resolution and low-computational-cost technique for an AT device.
We estimated the number of propagation modes using information theoretic criteria and the DL
technique. We proposed a method to estimate the optimal DL value required for guided wave estimation.
The proposed method did not involve processes of high computational cost. The proposed method was
evaluated experimentally using 4-mm-thick copper and bone-mimicking plates. The estimation error was
less than 4% and the calculation time when using the proposed method on a laptop computer was less
than 0.5 s. While the processor is different and is in fact faster than that which was used in the previous
conventional study, the calculation time is less than 1% of that of the conventional method. We believe
that our proposed method thus has the potential to accurately characterize the elastic properties of
cortical bone.
Acknowledgments: This work was supported in part by a Grant-in-Aid for Scientific Research (A) (Grant
No. 25249057), a Grant-in-Aid for JSPS Fellows (Grant No. 15J05687), and the PEPS program (15R03051
AMETCARMAT) from the Université Paris-Est. This project received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 Research and Innovation Program (Grant Agreement No. 682001,
project ERC Consolidator Grant 2015 BoneImplant). We thank David MacDonald, MSc, from Edanz Group
(www.edanzediting.com/ac) for editing a draft of this manuscript.
Author Contributions: Shigeaki Okumura analyzed the data and wrote the paper; Vu-Hieu Nguyen contributed
to the basic conception and discussion and gave feedback on the manuscript; Hirofumi Taki contributed to the
feedback on the signal processing algorithm and the manuscript; Guillaume Haïat contributed to the discussion
and gave feedback on the manuscript from an application perspective; Salah Naili contributed to the discussion
and provided feedback on the manuscript; Toru Sato contributed to the design of the signal processing algorithm
and gave feedback on the manuscript.
Conflicts of Interest: The authors declare no conflict of interest. The founding sponsors had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the
decision to publish the results.
265
Appl. Sci. 2018, 8, 652
Abbreviations
The following abbreviations are used in this manuscript:
QUS quantitative ultrasound

AT axial transmission
f –k frequency-wavenumber
SVD singular value decomposition
S-SVD sparse singular value decomposition
ESPRIT estimation of signal parameters via rotational invariance technique
SNR signal-to-noise ratio
MDL minimum description length
DL diagonal loading
RMSE root mean square error
References
1. Langton, C.; Palmer, S.; Porter, R. The measurement of broadband ultrasonic attenuation in cancellous bone.
Eng. Med. 1984, 13, 89–91. [CrossRef]
2. Mano, I.; Horii, K.; Takai, S.; Suzaki, T.; Nagaoka, H.; Otani, T. Development of novel ultrasonic bone
densitometry using acoustic parameters of cancellous bone for fast and slow waves. Jpn. J. Appl. Phys. 2006,
45, 4700–4702. [CrossRef]
3. Otani, T. Quantitative estimation of bone density and bone quality using acoustic parameters of cancellous
bone for fast and slow waves. Jpn. J. Appl. Phys. 2005, 44, 4578, [CrossRef]
4. Raum, K.; Grimal, Q.; Varga, P.; Barkmann, R.; Glüer, C.C.; Laugier, P. Ultrasound to assess bone quality.
Curr. Osteoporos. Rep. 2014, 12, 154–162. [CrossRef]
5. Ta, D.; Liu, C. Ultrasonic backscatter theory, method, and diagnostic instrument for cancellous bone. J. Acoust.
Soc. Am. 2016, 140, 3190–3190. [CrossRef]
6. Bochud, N.; Vallet, Q.; Minonzio, J.G.; Laugier, P. Predicting bone strength with ultrasonic guided waves.
Sci. Rep. 2017, 7, 43628, [CrossRef]
7. Wear, K.A.; Nagaraja, S.; Dreher, M.L.; Sadoughi, S.; Zhu, S.; Keaveny, T.M. Relationships among ultrasonic
and mechanical properties of cancellous bone in human calcaneus in vitro. Bone 2017, 103, 93–101. [CrossRef]
8. Okada, S.; Kawano, A.; Oue, H.; Takeda, Y.; Yokoi, M.; Koretake, K.; Tsuga, K. Preoperative evaluation of
bone quality for dental implantation using an ultrasound axial transmission device in an ex vivo model.
Clin. Exp. Dent. Res. 2017, 3, 81–86, [CrossRef]
9. Xu, K.; Ta, D.; Cassereau, D.; Hu, B.; Wang, W.; Laugier, P.; Minonzio, J.G. Multichannel processing for
dispersion curves extraction of ultrasonic axial-transmission signals: Comparisons and case studies. J. Acoust.
Soc. Am. 2016, 140, 1758–1770. [CrossRef]
10. Minonzio, J.G.; Talmant, M.; Laugier, P. Guided wave phase velocity measurement using multi-emitter
and multi-receiver arrays in the axial transmission configuration. J. Acoust. Soc. Am. 2010, 127, 2913–2919.
[CrossRef]
11. Minonzio, J.G.; Foiret, J.; Talmant, M.; Laugier, P. Impact of attenuation on guided mode wavenumber
measurement in axial transmission on bone mimicking plates. J. Acoust. Soc. Am. 2011, 130, 3574–3582.
[CrossRef]
12. Tran, T.N.; Nguyen, K.C.T.; Sacchi, M.D.; Le, L.H. Imaging ultrasonic dispersive guided wave energy in long
bones using linear Radon transform. Ultrasound Med. Biol. 2014, 40, 2715–2727. [CrossRef]
13. Tran, T.N.; Le, L.H.; Sacchi, M.D.; Nguyen, V.H.; Lou, E.H. Multichannel filtering and reconstruction of
ultrasonic guided wave fields using time intercept-slowness transform. J. Acoust. Soc. Am. 2014, 136, 248–259.
[CrossRef]
14. Dong, X.N.; Guo, X.E. The dependence of transversely isotropic elasticity of human femoral cortical bone on
porosity. J. Biomech. 2004, 37, 1281–1287. [CrossRef]
15. Naili, S.; Vu, M.B.; Grimal, Q.; Talmant, M.; Desceliers, C.; Soize, C.; Haïat, G. Influence of viscoelastic
and viscous absorption on ultrasonic wave propagation in cortical bone: Application to axial transmission.
J. Acoust. Soc. Am. 2010, 127, 2622–2634. [CrossRef]
266
Appl. Sci. 2018, 8, 652
16. Haïat, G.; Naili, S.; Grimal, Q.; Talmant, M.; Desceliers, C.; Soize, C. Influence of a gradient of material
properties on ultrasonic wave propagation in cortical bone: Application to axial transmission. J. Acoust.
Soc. Am. 2009, 125, 4043–4052. [CrossRef]
17. Nguyen, V.H.; Naili, S. Ultrasonic wave propagation in viscoelastic cortical bone plate coupled with fluids:
A spectral finite element study. Comput. Meth. Biomech. Biomed. Eng. 2013, 16, 963–974. [CrossRef]
18. Nguyen, V.H.; Tran, T.N.; Sacchi, M.D.; Naili, S.; Le, L.H. Computing dispersion curves of elastic/viscoelastic
transversely-isotropic bone plates coupled with soft tissue and marrow using semi-analytical finite element
(SAFE) method. Comput. Biol. Med. 2017, 87, 371–381. [CrossRef]
19. Sasso, M.; Haiat, G.; Talmant, M.; Laugier, P.; Naili, S. Singular value decomposition-based wave extraction
in axial transmission: Application to cortical bone ultrasonic characterization [correspondence]. IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 2008, 55, 1328–1332. [CrossRef]
20. Foiret, J.; Minonzio, J.G.; Chappard, C.; Talmant, M.; Laugier, P. Combined estimation of thickness and
velocities using ultrasound guided waves: A pioneering study on in vitro cortical bone samples. IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 2014, 61, 1478–1488. [CrossRef]
21. Xu, K.; Minonzio, J.G.; Ta, D.; Hu, B.; Wang, W.; Laugier, P. Sparse SVD method for high-resolution extraction
of the dispersion curves of ultrasonic guided waves. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2016,
63, 1514–1524. [CrossRef]
22. Okumura, S.; Nguyen, V.H.; Taki, H.; Haïat, G.; Naili, S.; Sato, T. Phase velocity estimation technique
based on adaptive beamforming for ultrasonic guided waves propagating along cortical long bones. Jpn. J.
Appl. Phys. 2017, 56, 07JF06. [CrossRef]
23. Wax, M.; Kailath, T. Detection of signals by information theoretic criteria. IEEE Trans. Acoust. Speech
Signal Process. 1985, 33, 387–392. [CrossRef]
24. Wax, M. Detection and localization of multiple sources via the stochastic signals model. IEEE Trans.
Signal Process. 1991, 39, 2450–2456. [CrossRef]
25. Lombardini, F.; Gini, F. Model order selection in multi-baseline interferometric radar systems. EURASIP J.
Adv. Signal Process. 2005, 2005, 108784. [CrossRef]
26. Huang, Y.; Ferro-Famil, L.; Reigber, A. Under-foliage object imaging using SAR tomography and polarimetric
spectral estimators. IEEE Trans. Geosci. Remote Sens. 2012, 50, 2213–2225. [CrossRef]
27. Sauer, S.; Ferro-Famil, L.; Reigber, A.; Pottier, E. Physical parameter extraction over urban areas using L-band
POLSAR data and interferometric baseline diversity. In Proceedings of the 2007 IEEE International IGARSS
Geoscience and Remote Sensing Symposium, Barcelona, Spain, 23–28 July 2007; pp. 5045–5048.
28. Takao, K.; Kikuma, N. An adaptive array utilizing an adaptive spatial averaging technique for multipath
environments. IEEE Trans. Antennas Propag. 1987, 35, 1389–1396. [CrossRef]
29. Shan, T.J.; Wax, M.; Kailath, T. On spatial smoothing for direction-of-arrival estimation of coherent signals.
IEEE Trans. Acoust. Speech Signal Process. 1985, 33, 806–811. [CrossRef]
30. Saarnisaari, H. TLS-ESPRIT in a time delay estimation. In Proceedings of the 1997 IEEE 47th Vehicular
Technology Conference, Phoenix, AZ, USA, 4–7 May 1997; Volume 3, pp. 1619–1623.
c 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
267
applied
sciences
Article
Forward and Inverse Studies on Scattering
of Rayleigh Wave at Surface Flaws
Bin Wang, Yihui Da and Zhenghua Qian *
State Key Laboratory of Mechanics and Control of Mechanical Structures/College of Aerospace Engineering,
Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China; wangbin1982@nuaa.edu.cn (B.W.);
dayihui@163.com (Y.D.)
* Correspondence: qianzh@nuaa.edu.cn; Tel.: +86-25-8489-5952
Featured Application: Non-destructive evaluation.
Abstract: The Rayleigh wave has been frequently applied in geological seismic inspection and
ultrasonic non-destructive testing, due to its low attenuation and dispersion. A thorough and
effective utilization of Rayleigh wave requires better understanding of its scattering phenomenon.
The paper analyzes the scattering of Rayleigh wave at the canyon-shaped flaws on the surface,
both in forward and inverse aspects. Firstly, we suggest a modified boundary element method (BEM)
incorporating the far-field displacement patterns into the traditional BEM equation set. Results
show that the modified BEM is an efficient and accurate approach for calculating far-field reflection
coefficients. Secondly, we propose an inverse reconstruction procedure for the flaw shape using
reflection coefficients of Rayleigh wave. By theoretical deduction, it can be proved that the objective
function of flaw depth d(x1 ) is approximately expressed as an inverse Fourier transform of reflection
coefficients in wavenumber domain. Numerical examples are given by substituting the reflection
coefficients obtained from the forward analysis into the inversion algorithm, and good agreements are
shown between the reconstructed flaw images and the geometric characteristics of the actual flaws.
Keywords: surface flaw; Rayleigh wave; scattering; modified BEM; reconstruction
1. Introduction
The non-destructive testing and structural health monitoring are of great importance in civil,
mechanical, and aerospace engineering industries. Corrosions, pittings, and cracks may occur
during service years of structures, and cause threats to public life and possession safety. Ultrasonic
inspection is one of the fundamental techniques for locating surface and hidden flaws. Compared
with traditionally-applied bulk waves, the Rayleigh surface wave has low attenuation and high energy
concentration, which makes it suitable for inspection of defects or abnormalities on or close to surface.
On the other hand, Rayleigh waves are not dispersive, as compared with other guided waves such as
Lamb, SH, and Love waves, thus easier for signal processing [1,2].
Due to the great potential of Rayleigh waves in NDT (Non-Destrctive Testing) applications, it is
necessary to examine their scattering and diffraction behaviors, both in the forward and inverse
aspects. Researchers have long been interested in forward calculation of scattering phenomena
for Rayleigh waves. One of pioneering work can be attributed to Ayster and Auld’s paper of
calculating the scattered field of Rayleigh waves by surface crack [3], which is followed by Achenbach,
who discussed the emission of surface waves from a crack by cyclic loading in detail [4]. Furthermore,
Lee and Liu deduced the analytical solution of a scattered Rayleigh wave field by semi-circular
canyons [5]. However, theoretical calculations are only applicable for special cases. For more general
situations, numerical approaches are required. As a widely applied numerical method, finite element

Appl. Sci. 2018, 8, 427
method (FEM) has been applied for Rayleigh wave field simulation as it passes through cracks [6],
slots [7], and notches [8] on the flat surface (of a large yet finite solid block). The FEM mesh was
performed within a whole elastic solid block [7,8] or a confined region of an infinite body with
absorption boundaries [6].
Although FEM is a sophisticated numerical approach, it is time- and memory-consuming for
scattering problems in a half-space, especially when far-field behaviors need to be examined. Thus,
some researchers turned to the boundary element method (BEM), in which only the boundaries need to
be meshed. Their early works include Wong [9] and Kawase’s [10] applications of discrete wavenumber
BEM for calculation of scattered Rayleigh wave fields in frequency domains. Sanchez-Sesma et al. [11]
and Ávila-Carrera et al. [12] used indirect BEM to solve similar problems. For half-space problems,
there have been two main modeling patterns: (i) analyzing and meshing the flaw surface only,
with usage of half-space Green’s functions, in which the interface boundary conditions are satisfied
automatically; (ii) analyzing and meshing the flaw surface and the whole interface, with usage of
full-space Green’s functions. For the former model, the element is fewer, however, the format of
Green’s functions tends to be complicated, usually not in a closed form, and requires a large amount of
resources for contour or numerical integration over complex wavenumber domains [10,13]. On the
other hand, for the latter model, although the whole interface is incorporated, the Green’s functions
are simpler and easier to apply in the BEM calculation. Thus, many recent researchers used the latter
modeling approach to study the scattering of Rayleigh and Lamb waves [14–17]. But researchers often
encounter the problem of “spurious reflection” because of the model truncation at far ends [18,19].
Arias [18] suggested a modified BEM method in which a far-field displacement pattern had been
introduced and substituted into the BEM equation set. By enforcing displacement and stress continuous
relations, the researchers were able to solve the scattering wave field while suppressing spurious
reflections. However, Arias’ method might not be accurate for calculating far-field reflection and
transmission coefficients. In this paper, we will propose an improved modified BEM method.
The inverse analysis is an important aspect and fundamental aim for studying the scattering
of Rayleigh wave. Many researchers have tried to relate the amplitude and time-of-flight (TOF)
of reflected wave with surface-breaking cracks [20–23] or notches [24,25], both in experimental and
numerical aspects, in order to achieve flaw location and sizing effect. These researchers mainly adopted
pitch-catch configurations, and were able to detect the positions and approximate sizes of defects,
but lacked further information, such as depth, width, and severity.
In order to obtain more information about defects and abnormalities in the near-surface layer,
geophysical scientists and engineers have tried tomographic reconstruction approaches using data from
scattered Rayleigh wave fields. Xia et al. [26] firstly used Rayleigh wave phase velocities at different
frequencies to reconstruct the S-wave velocity profile of multiple elastic layers on top of the half-space.
Furthermore, Xia et al. [27] and Shao et al. [28] extended the tomographic approach to the detection of
near-surface shallow cavities. Köhn et al. [29] used a similar inversion method to reconstruct surface
layer stiffness of an ancient sandstone façade. Chen et al. [30] adopted an inversion procedure to
detect residual stresses from general Rayleigh wave dispersion of a weakly anisotropic media. These
methods stem from wave signal processing theories (in earlier works [27,28]) or discretized equations
of motion (in later works [29,30]). Through the formulation of an “operator matrix” of inverse problem,
the researchers try to find an optimum solution of the “model vector” that is feasible to the “data set”.
Due to strong nonlinearity of general inverse problems, one needs to perform numerous iterations to
achieve an accurate solution.
In this paper, we will propose a convenient inverse method to reconstruct the location and
geometric properties of surface-opening flaws. The method is subsequent research for inversion
method for plate-thinning reconstruction using SH-waves [31,32] and Lamb waves [33]. We make
use of the reflection coefficients of Rayleigh waves obtained from the forward analysis in different
frequencies as the input data. These data can also be attained by experiments. By introduction
of Born approximation and far-field expressions of half-space Green’s functions, and the usage of
270
Appl. Sci. 2018, 8, 427
boundary integral equation over the flaw area, it can be proved that the objective function of flaw
depth d(x1 ) is approximately expressed as an inverse spatial Fourier transform of reflection coefficients
in wavenumber domain. This inversion procedure is especially suitable for reconstruction of weak
and shallow notches, and can also be used as an initial-step image for deeper and steeper canyons.
The paper will be organized as follows: In Section 2, we will discuss about the improved forward
problem using modified BEM in Section 2.1, and give some numerical examples and compare with
existing results in Section 2.2. In Section 3, we will elaborate on the mathematical formula for inverse
reconstruction method for surface-breaking canyon-shaped flaws in Section 3.1, and parametric
influences will be discussed in detail in Section 3.2. We will give conclusions in Section 4.
2. Forward Problem by Modified BEM
2.1. Formulation of Modified BEM

The problem to be solved is shown in Figure 1. An artificial canyon-shaped flaw has been cut on
the surface of an isotropic elastic half-space. The incident Rayleigh wave propagates from left to the
right side, and the reflected wave is observed at the far end of the left side.
For accurately computing the scattering wave field, we divide the forward analysis into two major
steps, as shown in Figure 2.
(i) A time-harmonic incident Rayleigh wave field in the intact half-space is considered, and the
correspondent stress σijinc is calculated along the curve Γ1− , which is actually the flaw surface.
(ii) Since the actual traction on Γ1− is zero, the stress σijinc is released by exerting a compensation
traction tsca
i = −σijinc n j along Γ1− . The scattering problem is transformed into the question of solving
the wave field in the flawed half-space subjected to load tsca i .
Figure 1. Problem configuration.
Figure 2. Two steps of calculating scattering wave field, (a) Step 1; (b) Step 2.
271
Appl. Sci. 2018, 8, 427
Then, the displacement in frequency domain of scattering wave field at an arbitrary point can be
expressed in the form of boundary integral equation (BIE), as shown by Equation (1) [34].
' , -
i ( X, Ω ) =
c pi ( X )usca i ( x, Ω ) − Tip ( x, X, Ω ) ui ( x, Ω ) dΓ ( x )(i, p = 1, 2)
Gip ( x, X, Ω)tsca sca
(1)
Γ
i and ti are displacement and boundary traction, respectively. Gip ( x, X, Ω ) and Tip ( x, X, Ω )
where usca sca
are fundamental solutions representing the i-th component of the displacement and stress at point x
due to a unit time-harmonic concentrated load of angular frequency Ω exerted in the p-th direction
at point X in the elastic full-space. The coefficient c pi ( X ) = δpi for X ∈ V, c pi ( X ) = δpi /2 for point
X ∈ Γ which is smooth locally, c pi ( X ) = 0 for X ∈
/ V ∪ Γ. It should be noted that here the boundary Γ
includes both the flaw area and the intact infinite surface, as the full-space fundamental solutions have
been chosen.
Now the whole surface is divided into six parts, Γ− +
∞ and Γ∞ tending to infinity on the left and
right side, respectively; Γ0− and Γ0+ representing the near-field boundaries; Γ1− and Γ1+ the left and
right half of the flaw surface, respectively. Consequently, the scattered wave field is rewritten from
Equation (1):
. .
i ( X, Ω ) = − Γ−
c pi ( X )usca + Tip ( x, X, Ω ) ui ( x, Ω ) dΓ ( x ) − Γ− ∪ Γ+ Tip ( x, X, Ω ) ui ( x, Ω ) dΓ ( x )
sca sca
. ∞ ∪Γ∞ 0 0 (2)
+ Γ− ∪Γ+ [ Gip ( x, X, Ω)tsca
i ( x, Ω ) − T ip ( x, X, Ω ) u sca ( x, Ω )] dΓ ( x ), (i, p = 1, 2)
i
1 1
where the traction-free boundary condition on Γ0− ∪ Γ0+ ∪ Γ− +

∞ ∪ Γ∞ has been taken into account.
For traditional BEM, only the near-field boundaries Γ0 ∪ Γ0 ∪ Γ1− ∪ Γ1+ are meshed, thus, the first
− +
item on the right hand side of Equation (2) is omitted. This treatment can cause spurious reflected
waves at the truncation points which do not attenuate in the 2D half-space, which are unwanted.
It is noticed that, if the truncation point is far enough from the flaw area, the displacement fields
on the infinite boundaries will follow a Rayleigh wave pattern travelling out of the flaw region [18],
as expressed by
− i− ( x, Ω), x ∈ Γ−
i ( x, Ω ) ≈ R ( Ω ) u
usca ∞
(3)
ui ( x, Ω) ≈ R (Ω)ui+ ( x, Ω), x ∈ Γ+
sca +
∞
where ui± . is the displacement field of the unitary Rayleigh wave travelling in positive (+) or negative
(−) directions, and R± the coefficient to be solved. Equation (3) makes it possible to take the far-field
integrals in Equation (2) into consideration. Then, for any point on the boundary which is locally
smooth, we have
.
2 u p ( X, Ω ) +
1 sca
R− (Ω) A− + +
p ( X, Ω ) + R ( Ω ) A p ( X, Ω ) + Γ− ∪ Γ+ ∪ Γ− ∪ Γ+ Tip ( x, X, Ω ) ui ( x, Ω ) dΓ ( x )
sca
. 0 0 1 1 (4)
= Γ− ∪ Γ+ Gip ( x, X, Ω)tsca
i ( x, Ω ) dΓ ( x ) , ( i, p = 1, 2 ) .
0 0
.
where the correction terms A± −
p having the definition A p ( X, Ω ) = Γ− Tip ( x, X, Ω)ui− ( x, Ω)dΓ( x ) and
. +
∞
A p ( X, Ω) = Γ+ Tip ( x, X, Ω)ui ( x, Ω)dΓ( x ). They are constants. In paper [18], the coefficients R± are
+
∞
not explicit in the modified BEM formulation, consequently, the reflection coefficients obtained from x1
and x2 components of displacement might not coincide with each other. As an improvement, we will
try to solve R± together with displacement fields.
The correction terms A± p involve the integral over an infinite boundary, however, by applying the
reciprocity theorem, the integral can be transferred to a bounded region [18,35].
.
A+ i+ ( X, Ω) −
p ( X, Ω ) = c pi ( X ) u Tip ( x, X, Ω)ui+ ( x, Ω)dΓ( x )
Γ0+ (5)
.
+ + Tip ( x, X, Ω)ui+ ( x, Ω)]dΓ( x )(i, p = 1, 2)
Γ1+ ∪ Γ2 [ Gip ( x, X, Ω ) ti ( x, Ω ) −
272
Appl. Sci. 2018, 8, 427
.
A− i− ( X, Ω) − Γ+ Tip ( x, X, Ω)ui− ( x, Ω)dΓ( x )
p ( X, Ω ) = c pi ( X ) u
. , 0 - (6)
+ Γ+ ∪ Γ Gip ( x, X, Ω)ti− ( x, Ω) − Tip ( x, X, Ω)ui− ( x, Ω) dΓ( x )(i, p = 1, 2)
2
1
where Γ2 represents an auxiliary vertical boundary of 2–3 Rayleigh wavelengths, see Figure 2.
t+ = − +
σi1 n1 , and ti− = −
σi1 n1 are defined on Γ2 with n1 the unit normal vector pointing in positive
i
±
x1-direction. σi1 is the stress field of the unitary Rayleigh wave propagating in positive or negative
x1-directions, respectively.
By substitution of Equations (5) and (6) into Equation (4), we can obtain an equation set with
the displacement and coefficients R± as the undetermined variables. Here, we adopt the constant
elements (see Figure 3) in the BEM discretization, in which one element is represented by the only
node in the middle, and obtain the matrix representation in the form of
[T]{usca } = [G]{tsca } (7)
Here, we assume that for the 1st or the N-th node, the displacements are not independent,
and can be expressed as R− (Ω)ui− ( x1 , Ω) and R+ (Ω)ui+ ( x N , Ω), respectively, where x1 and x N are
coordinates of the 1st and the N-th node, respectively. Thus, the submatrices concerning the 1st and
the N-th elements in [T] are combined with the correction terms A± p , and placed as the first and last
columns. Consequently, the vector {usca } contains the displacements of nodes 2~(N − 1), together
with coefficients R± , as shown by
T
{usca } = R− u21 u22 ... u P1 u P2 ... u ( N −1)1 u ( N −1)2 R+ (8)
where u Pi represents the displacement in ith direction at the Pth node.

The matrix [T] is expressed in detail as
⎡ − + ⎤
S11 T2111 T2211 ... Tp111 Tp211 ... T( N −1)111 T( N −1)211 S11
⎢ −
S12 T2112 T2212 ... Tp112 Tp112 ... T( N −1)112 T( N −1)212 +
S12 ⎥
⎢ ⎥
⎢ .. .. .. .. .. .. .. .. ⎥
⎢ ⎥
⎢ . . . . . . . . ⎥
⎢ ⎥
⎢ S− T21j1 T22j1 ... Tp1j1 Tp2j1 ... T( N −1)1j1 T( N −1)2j1 S+ ⎥
[T] = ⎢
⎢
j1 j1 ⎥
⎥ (9)
⎢ S−j2 T21j2 T22j2 ... Tp1j2 Tp1j2 ... T( N −1)1j2 T( N −1)2j2 S+j2 ⎥
⎢ ⎥
⎢ .. .. .. .. .. .. .. .. ⎥
⎢ . . . . . . . . ⎥
⎢ ⎥
⎣ S−
N1 T21N1 T21N1 ... Tp1N1 Tp2N1 ... T( N −1)1N1 T( N −1)2N1 S+
N1
⎦
S−
N2 T21N1 T21N1 ... Tp1N2 Tp1N2 ... T( N −1)1N2 T( N −1)2N2 S+
N2
The elements in the first and last columns are defined as
S− −
J1 = A J1 + T11J1 u
−
11 − −
+ T12J1 u12 ,S J2 = A−J2 + T11J2 u
−
11 −
+ T12J2 u12
(10)
S+
J1 = T
u +
N1J1 N1 + T
u
N2J1 N2
+
+ A + +
, S
J1 J2 = T
u
N1J2 N1
+
+ T
u +
N2J2 N2 + A+J2

in which, A± Ji = A± x J , obtained from Equations (5) and (6), and the elements TPiJk = (1/2)δPj δik +
. i
Γ P Tik x, x J , Ω dΓ, (i, k = 1, 2). The size of matrix [T] is 2N × 2(N − 1), and {u } 2(N − 1) × 1.
sca
On the other hand, [G] is a 2N × 2N0 sized matrix, {tsca } is a 2N0 × 1 sized vector, where N0
is the number of nodes on which the tractions are nonzero. The elements in [G] can be written as
.
GPiJk = Γ Tik x, x J , Ω dΓ, and the elements in {tsca } is tsca Pi x, x J , Ω .
P
Since the matrix [T] is not square-sized, the equation is solved in the least-square sense to
achieve error minimization. In the final solution, we can obtain reflection coefficient R− at the
first and transmission coefficient R+ . at the last element simultaneously, as well as the whole
near-field displacements.
273
Appl. Sci. 2018, 8, 427
Figure 3. Boundary element method (BEM) discretization by constant elements.
2.2. Verification of Reflection Coefficients

For verification of the reflection coefficients obtained from Equation (7), we adopt the scattering
equation in half-space, which uses the half-space Green’s function to calculate the far-field wave
pattern, as shown in [36]:
' H ( x, X )
∂Gln
n (X) =
usca Cijkl n j ( x ) n ( x ) dΓ ( x ) (i, j, k, l = 1, 2)
uinc (11)
Γ1− ∪ Γ1+ ∂xk
where Cijkl is the stiffness tensor, n j is the unit normal vector pointing outward the flaw area,
H ( x, X ) represents the Green’s functions in the half-space. In the far-field, the half-space
and Gln
Green’s function can be expressed in the Rayleigh wave form [37]:
H
Gαβ ( x, X ) = f (ξ 0 ) Apα ( x2 ) Bp β ( X2 )e+iξ 0 ( x1 −X1 ) , (α, β = 1, 2). (12)
where
−1 (i = 2, x1 < X1 ) −1 ( j = 2, X1 < x1 )
A= ...B =
1 otherwise 1 otherwise
RT0 + RT0 x2 ξ 02 + R2T0 + RL0 x2
p1 ( x2 ) = ξ0 e − 2R e
L0 ξ 0
ξ 2 + R2 (13)
p2 ( x2 ) = −ie+ RT0 x2 + i 0 2ξ 2 T0 e+ RL0 x2
0
4iξ 4 R
f (ξ 0 ) = μk2 F0 (L0
ξ0 )
2 2T
F (ξ ) = ξ + R2T0 − 4ξ 2 RL0 RT0
in which kL = Ω/cL , kT = Ω/cT , ξ 0 are wavenumbers of S, P, and Rayleigh waves, respectively,

and R2L0 = ξ 02 − k2L , R2T0 = ξ 02 − k2T . The Rayleigh wavenumber ξ 0 satisfies F (ξ 0 ) = 0, and F (ξ 0 ) is the
derivative of F (ξ ) at ξ = ξ 0 .
The reflection coefficients derived from Equation (11) should coincide with that obtained directly from
Equation (7). A numerical example will be given to show the validity of reflection coefficients. Moreover,
the scattering Equation (11) is more useful in the inverse reconstruction procedure of the flaw shape.
2.3. Results of Forward Problem
Example 1. To verify the modified BEM method, we now try to solve the classical Rayleigh wave scattering
problem presented by Kawase [10], for which the solutions have already been obtained from discrete wavenumber
method applying half-space Green functions.
As shown by Figure 4a, an artificial half-circular canyon-shaped flaw is set on the surface of an
isotropic elastic half-space, with Poisson’s ratio of 1/3. The incident wave is a unitary Rayleigh wave
travelling from left to right, whose time-domain waveform is expressed as

u( T ) = 1/2Ω2 ( T − T0 )2 − 1 exp −(Ω/2)2 ( T − T0 )2 (14)
in which Ω = Ωr/cT is the normalized circular frequency, where r is the radius of canyon, and cT is
the shear wave velocity. T = tcT /r represents the normalized time array. The time-domain signal is
shown in Figure 5, when Ω = 2π and T0 = 2π.
274
Appl. Sci. 2018, 8, 427
Figure 4. Numerical examples: (a) a half-space with an arc-shaped canyon flaw; (b) a half-space with
double canyons.
Figure 5. Time-domain signal of the incident wave.
We perform a Fourier transform to the incident wave signal in time-domain to obtain its frequency
spectrum, and calculate the displacement field and reflection coefficients at each frequency point by the
modified BEM. As an example, the whole displacement field along the surface at frequency Ω = 2π is
plotted in Figure 6, and is compared with the results obtained by Kawase [10]. It is seen that the BEM
displacements show good agreement with the previous results.
Figure 6. Comparisons of the surface displacements obtained by the modified BEM and those in Kawase [10].
275
Appl. Sci. 2018, 8, 427
Displacement fields at other frequency points have also been calculated in the same manner.
An inverse Fourier transform is then applied to the frequency-domain displacements to attain
time-domain wave signals for different points along the surface, as shown in Figure 7a,b, from
which we can see that the spurious reflection waves are not generated.
Figure 7. Displacement at different surface points in time-domain (a) x1 component; (b) x2 component.
Example 2. Similarly with Example 1, an artificial arc-shaped canyon is located on the surface of an isotropic
elastic half-space, whose depth–radius ratio dmax /r = 0.2 . The Poisson’s ratio is 1/3. We calculated the
reflection coefficients in frequency domain within range Ω = Ωr/c T from 0.2 to 5.0 at an interval of 0.2.
The circular dots in Figure 8 shows the reflection coefficients obtained directly from the
modified BEM program, and the rectangular dots represents those derived from half-space scattering
Equation (11). The good agreement illustrates the validity of the forward analysis.
276
Appl. Sci. 2018, 8, 427
Figure 8. Reflection amplitudes obtained by BEM and half-space scattering equation respectively.
This example will be used in the following inverse reconstruction.
3. Inverse Problem by Modified BEM
3.1. Formulation of Inverse Reconstruction

We now try to reconstruct the position and geometric properties of the canyon-shaped flaw. In order to
quantitatively depict the flaw’s geometric characteristics, an objective function d(x1 ) is defined in x1 domain,
which means the flaw depth at coordinate x1, whereas in the intact area, its value is zero.
The scattering Equation (11) in half-space is adopted for inversion. In this paper, a shallow flaw
is considered, which can be thought of as a weak scatterer. In this case, the Born approximation [36]
can be used for linearization, which replaces the total wave field by the incident wave field on flaw
1 ≈ u1 . Then Equation (11) becomes
boundaries. e.g., utot inc
' H ( x, X )
∂Gln
n (X) ≈
usca Cijkl n j ( x ) n ( x ) dΓ ( x ) (i, j, k, l, n = 1, 2)
uinc (15)
Γ1− ∪ Γ1+ ∂xk
It is noticed that, on the original half-space surface Γ1 (see the dotted line in Figure 1), the traction
induced by half-space Green functions is zero. Thus, the surface for integration can be extended to
Γ1− ∪ Γ1+ ∪ Γ1 . Furthermore, after substitution of Lamé constant λ and shear modulus μ, the scattering
equation is expanded as

. ∂GH H
∂G1n H
inc ( λ + 2μ ) + n ∂G1n uinc + n ∂G2n uinc λ
H
n ( X ) ≈ Γ− ∪ Γ+ ∪ Γ1
usca 1 + n2 ∂x2 u2
n1 ∂x1n uinc 2 ∂x1 2 1 ∂x1 1
1 1 1
! (16)
∂GH H
∂G1n H
∂G1n H
∂G2n
+ n2 ∂x2n uinc
1 + n1 ∂x u2 + n2 ∂x u1 + n1 ∂x u2
1
inc
2
inc inc μ dΓ ( x ) (i, j, k, l, n = 1, 2)
2 1
Since surface Γ1− ∪ Γ1+ ∪ Γ1 is closed, we can perform Gauss’s theorem to the flawed region
and obtain

. ∂2 G1n
H H ∂uinc
∂G1n ∂2 G2n
H H ∂uinc
∂G2n
n (X) ≈
usca uinc + + uinc + (λ + 2μ)
1 2
V ∂x12 1 ∂x1 ∂x1 ∂x22 2 ∂x2 ∂x2

∂2 G1n
H H ∂uinc
∂G1n ∂2 G2n
H H ∂uinc
∂G2n
+ inc
∂x1 ∂x2 u2 + 2
∂x1 ∂x2 + inc
∂x1 ∂x2 u1 + 1
∂x2 ∂x1 λ
(17)
∂2 G2n
H H ∂uinc
∂G2n ∂2 G1n
H H ∂uinc
∂G1n
+ inc
∂x1 ∂x2 u1 + 1
∂x1 ∂x2 + inc
∂x1 ∂x2 u2 + 2
∂x2 ∂x1
!
∂2 G1n
H H ∂uinc
∂G1n ∂2 G2n
H H ∂uinc
∂G2n
+ ∂x22
uinc
1 + 1
∂x2 ∂x2 + ∂x12
uinc
2 + 2
∂x1 ∂x1 μ dΓ( x ) (i, j, k, l, n = 1, 2)
It is assumed that a pure incident Rayleigh wave mode is sent from left to right, and the reflected wave
is observed at the far field on the left side. Thus, the incident wave with wavenumber ξ0 can be expressed as
+iξ 0 x1
1 ( x, ξ 0 ) = A ( ξ 0 ) p1 ( x2 ) e
uinc inc
inc ( ξ ) p ( x ) e+iξ 0 x1 (18)
uinc
2 ( x, ξ 0 ) = A 0 2 2
277
Appl. Sci. 2018, 8, 427
where pi ( x2 ) (i = 1, 2) is defined in Equation (13), and Ainc represents the complex amplitude of the
incident wave.
We use the far-field form of half-space Green’s functions (12), and substitute them into
Equation (17), and obtain the reflected wave of the same wave number ξ 0 , as shown by Equation (19):
2
.
−2ξ 02 p21 ( x2 ) + d pdx2 (2x2 ) p2 ( x2 ) + dpdx
2 ( x2 )
2
n ( X, ξ 0 ) ≈ A ( ξ 0 ) V
usca (λ + 2μ)
inc
2 2
dp1 ( x2 ) dp2 ( x2 )
+ iξ 0 dx2 p2 ( x2 ) (λ + 3μ) + iξ 0 dx2 p1 ( x2 ) (3λ + μ)
2 ! (19)
+ −2ξ 0 p2 ( x2 ) + d pdx1 (2x2 ) p1 ( x2 ) + dpdx 1 ( x2 )
2
2 2 μ ·e+iξ 0 ·2x1 dV · f (ξ 0 ) · pn ( X2 )e−iξ 0 · X1
2 2
(i, j, k, l, n = 1, 2)
If one takes a thorough observation on Equation (19), he can find that pn ( X2 ) exp(−iξ 0 · X1 )
represents a unitary reflected Rayleigh wave travelling from right to left, and the integral over V
is taken with respect to coordinate x of field points, and can be viewed as a complex amplitude of
that unitary reflected Rayleigh wave. Also, the integrand is separated into the part in the bracket
concerning x2 coordinate, and an exponential item exp(+iξ 0 ·2x1 ).
This inspires us to rewrite the expression of reflected wave into the following form
.
n ( X, ξ 0 ) ≈ A ( ξ 0 ) f ( ξ 0 )· V Fun( x2 , ξ 0 ) exp(+iξ 0 ·2x1 ) dV · pn ( X2 ) exp(−iξ 0 · X1 )
usca inc
(20)
≡ Aref (ξ 0 ) pn ( X2 ) exp(−iξ 0 · X1 )
where the Fun( x2 , ξ 0 ) is the expression in the bracket in Equation (19), and the complex amplitude
for the reflected wave is termed as Aref . Now we define the reflection coefficient Cref (ξ 0 ) as the ratio
between Aref (ξ 0 ) and Ainc (ξ 0 ), so that
'
Cref (ξ 0 ) ≈ f (ξ 0 )· Fun( x2 , ξ 0 ) exp(+iξ 0 ·2x1 )dV (21)
V
Also, the volume integral is further rewritten as a dual integral:
' ' ∞ ' 0
Fun( x2 , ξ 0 ) exp(+iξ 0 ·2x1 )dV = exp(+iξ 0 ·2x1 )dx1 Fun( x2 , ξ 0 )dx2 (22)
V −∞ − d ( x1 )
where d( x1 ) is the flaw depth at coordinate x1 . The integral range has been extended from minus
infinity to infinity, since d( x1 ) = 0 in the intact areas.
After organization, the reflection coefficient is expressed as
.∞ .0
Cref (ξ 0 ) ≈ −∞ e+iξ 0 ·2x1 dx1 − d ( x1 ) cTT (ξ 0 )e+2RT0 x2 + cTL (ξ 0 )e+2( RT0 + RL0 ) x2 + cLL (ξ 0 )e+2RL0 x2 dx2 (23)
where cTT , cTL , and cLL are functions of wavenumber ξ 0 , and are expressed in detail as
k4
cTT (ξ 0 ) = 2μ ξT02
ξ 02 + R2T0 2 2
0
cTL (ξ 0 ) = 2ξ 2 R λkL0 ( RT0 − RL0 ) + μ 7ξ 0 RT0 − 5ξ 02 RL0 + 4R3L0 − R3T0 − RL0 R2T0 + 6R2L0 RT0 (24)
0 L0
2
k 4 ( ξ 2 + R2 )
cLL (ξ 0 ) = −(λ + 2μ) L0 2ξ04 R2 T0
0 L0
If we perform an integral over x2 first, we can rewrite Equation (23) as
.∞ −2R d ( x1 ) −( RT0 + RL0 )d( x1 ) −2R d ( x1 )

Cref (ξ 0 ) ≈ −∞ cTT (ξ 0 ) 1−e 2RT0
T0
+ cTL (ξ 0 ) 1−e RT0 + RL0 + cLL (ξ 0 ) 1−e 2RL0
L0
e+iξ 0 ·2x1 dx1 (25)
Since the flaw depth is considered small, it is natural to apply Taylor expansion for the exponential
functions in Equation (25) at d( x1 ) = 0, i.e.,
278
Appl. Sci. 2018, 8, 427
' ∞
Cref (ξ 0 ) ≈ [cTT (ξ 0 ) + cTL (ξ 0 ) + cLL (ξ 0 )]d( x1 )e+iξ 0 ·2x1 dx1 (26)
−∞
Then, it is found that the objective function d( x1 ) can be reconstructed using a spatial inverse
Fourier transform, as illustrated by
' ∞
1 Cref (ξ 0 )
d ( x1 ) ≈ e−ix1 ·2ξ 0 d(2ξ 0 ) (27)
2π −∞ cTT (ξ 0 ) + cTL (ξ 0 ) + cLL (ξ 0 )
It should be noted that the inversion formula adopts a series of assumptions for linearization.
Thus, parametric analysis should be performed to show the validity range of this method.
3.2. Numerical Examples

In this sub-section, numerical examples are given to show the validity of the inversion procedure.
Also, parametric studies are performed to illustrate which kind of flaw is the most suitable for this
reconstruction method.
Here, we mainly focus on the arc-shaped canyons as representatives of surface flaws. It is believed
that other shapes can lead to similar conclusions.
3.2.1. Choice of Parameters

The radius r, maximum depth dmax in Figure 4a, as well as the distance l in Figure Figure 4b
for the double-flaw case, are the key geometric parameters for arc-shaped canyons; the maximum
frequency Ωmax is an important parameter of input data.
For generality, we perform dimensionless analysis in this paper. The normalized radius R = r/r
is kept constant as 1; the depth is normalized as D = d/r; and frequency Ω = Ωr/cT , where cT
is the shear wave velocity in the half-space. Meanwhile, the normalized coordinates are written as
Xi = xi /r (i = 1, 2.)
The readers can readily verify that, in dimensionless analysis, the model (i) with radius r,
maximum depth dmax , and frequency range 0 ∼ Ωmax , is equivalent to model (ii) with radius
2r, maximum depth 2dmax , and frequency range 0 ∼ 0.5Ωmax . Thus, the absolute value of radius is of
little importance. Instead, the depth–radius ratio dmax /r (or Dmax /R) is a more essential parameter
showing the “steepness” of the canyon. Another important factor is the normalized frequency range,
which reflects the incident Rayleigh wavelength in comparison with the geometric flaw scale.
3.2.2. Reconstruction Results for Models of Different Depth–Radius Ratio

We established three canyon models with depth–radius ratios dmax /r equaling 0.1, 0.2, and 0.5,
respectively. For each model, we calculated the reflection coefficients of Rayleigh wave within the
frequency range from 0 to Ωmax = 5.0, with a uniform interval of dΩ = 0.2, as can be illustrated in
Figure 8. These coefficients are substituted into the inversion Equation (27) to obtain the flaw depth
function d( x1 ).
It should be noted, Equation (27) takes an integral from minus to positive infinity. However, the
reflection coefficients are only available at positive wavenumber points. Since d( x1 ) is a real function,
we can define Cref (−ξ 0 ) as the complex conjugate of Cref (ξ 0 ) (for ξ 0 > 0), i.e., Cref (−ξ 0 ) = Cref∗ (ξ 0 ).
The functions cTT (ξ 0 ), cTL (ξ 0 ), and cLL (ξ 0 ) are also defined similarly. The whole integrand function is
defined as 0 at ξ 0 = 0. Moreover, the integration range is reduced to 0 ∼ Ωmax .
After discretization, the integral of Equation (27) becomes

N −1
Cref,k exp(−2iπpk/N ) Cref,k∗ exp(+2iπpk/N )
d( pΔx1 ) = ∑ k + ck + ck
cTT
+ k∗ + ck∗ + ck∗
cTT
·2Δξ 0 (28)
k =0 TL LL TL LL
279
Appl. Sci. 2018, 8, 427
where Δx1 = π/NΔξ 0 is the increment in x1 coordinate. Integers p and k are indices in spatial and
wavenumber series, respectively. The discrete inverse Fourier transform is easily performed by FFT
(Fast Fourier Transform) algorithm.
The results are plotted in Figure 9. It is observed that when ratio dmax /r is as small as 0.1 (see
Figure 9a), the geometric properties (such as location, depth, and width) and shape of the flaw can be
reconstructed with good precision. When dmax /r value increases to 0.5 (see Figure 9c), the location of
flaw can still be precisely detected, and the width and depth are slightly larger than the reality. However,
the reconstructed shape is not as accurate as its shallower counterparts. The image underestimates
the bottom region, and looks as if there were two local maximum depth points. The reason why
the shallow and flat flaws are reconstructed better can be attributed to the application range of Born
approximation, which is especially suitable for weak scatterers.
Figure 9. Inverse reconstruction results of surface canyon-shaped flaws with different depth–radius
ratios (a) dmax /r = 0.1; (b) dmax /r = 0.2; (c) dmax /r = 0.5.
Even the reconstructed image is not accurate for deeper and steeper canyons, the geometric
characteristics (e.g., location, depth, and width) can be obtained with acceptable accuracy, so the image
can be used as the “initial solution” in the following iterative reconstruction algorithm.
280
Appl. Sci. 2018, 8, 427
3.2.3. Reconstruction Results Using Different Frequency Range

It is necessary to discuss how the frequency range affects the inverse reconstruction results,
which helps us understand the proper choice of incident wavelength in experiments. We have
recalculated the inverse reconstruction for the flaws with dmax /r equaling 0.2 and 0.5, using frequency
range Ω = 0 ∼ 5.0, 0 ∼ 2.0, and 0 ∼ 1.0, respectively.
The reconstruction results are plotted in Figure 10. For the case of dmax /r = 0.2 (see Figure 10a),
the frequency range of Ω = 0 ∼ 1.0 (see the dash-dot curve) only gives a vague image, yet it still
obtains the correct location. When the frequency range increases to Ω = 0 ∼ 2.0, the flaw shape seems
to agree well with the reality, however, the image underestimates the flaw depth and overestimates the
width. As the full data of reflection coefficients with Ω = 0 ∼ 5.0 are used in the inversion process,
the flaw image becomes clearer, and the width and depth are more accurate, although the image for
part of the bottom is deviated from reality.
Figure 10. Inverse reconstruction results of surface canyon-shaped flaws with different frequency
ranges (a) dmax /r = 0.2; (b) dmax /r = 0.5.
For the case of dmax /r = 0.5 (see Figure 10b), if only the low frequency parts of reflection data
(e.g., Ω = 0 ∼ 1.0 and Ω = 0 ∼ 2.0) are taken into the inversion process, the results imply the
existence of the flaw but the images are very blurred. It is only when higher frequency portions
are taken into consideration that the flaw image becomes clearer and width and depth are much
more accurate.
From these two examples, it can be concluded that the low-frequency or long-wavelength
components of the reflection signals (e.g., Ω = 0 ∼ 1.0) control the canyon’s basic information,
such as existence and location; while the usage of higher frequency components (e.g., Ω = 1.0 ∼ 5.0)
increases the clarity of image and gives more detailed geometric information, such as depth and
281
Appl. Sci. 2018, 8, 427
width. It should be noted that the effect of high frequency components of enhancing image resolution
is limited by Born approximation. As the Rayleigh wave becomes shorter in wavelength, it will
penetrate less deeply into the elastic half-space, thus, the canyon becomes a large scatterer, and Born
approximation loses validity. Actually, in the previous cases, if the high frequency components for
Ω > 5.0 are taken into consideration, the reconstructed images change little from the ones obtained
from data with Ω = 1.0 ∼ 5.0.
3.2.4. Reconstruction Results of Double-Canyon Type Flaws

In order to examine the validity of our inversion approach in multiple flaw cases as seen in
Figure 4b, we introduce here two models with double-canyon type flaws. For each model, two canyons
with different geometric parameters are cut on the surface of the half-space with a distance l in between.
The geometric parameters are set as (i) normalized radius: R1 = R2 = 1; (ii) normalized depth:
dmax,1 = 0.2, dmax,2 = 0.1; (iii) normalized edge-to-edge distance between two flaws: 1 L = l/r = 2.0,
2 L = l/r = 0.2.
The inversion results for the two models are plotted in Figure 11, in which input data with
frequency ranges Ω = 0 ∼ 1.0, Ω = 0 ∼ 2.0, and Ω = 0 ∼ 5.0 are shown in different curve types.
It can be seen that, when the two canyons are distant (see Figure 11a), even the data with low frequency
range can give correct locations of both two flaws and rough estimations of their depth and width.
As the input frequency range increases to Ω = 0 ∼ 5.0, the geometric parameters of both two flaws
are reconstructed with better accuracy.
On the other hand, if the two flaws are close with each other (see Figure 11b), the vague
image reconstructed from low frequency range data cannot distinguish the two canyons. As the
frequency range increases, the resolution becomes higher, and the images of two canyons are clearly
separated. Thus, it is concluded that the frequency range Ω = 0 ∼ 5.0 is necessary for reconstruction
of multiple flaws.
Figure 11. Inverse reconstruction results of double-canyon flaws (a) l/r = 2.0; (b) l/r = 0.2.
282
Appl. Sci. 2018, 8, 427
4. Conclusions
In this paper we present the studies of forward and inverse analysis on Rayleigh wave scattering
by surface-breaking canyon-shaped flaws.
For the forward problem, we show that the modified boundary element method is an effective
numerical technique to calculate the scattered wave field. By introduction of modification items
concerning far-field displacement patterns in the BEM matrix, one can effectively suppress artificial
reflected waves induced by model truncation. In our improved modified BEM version, the far-field
reflection and transmission coefficients are solved simultaneously with the whole displacement field in
the least-square sense. The improved version corrects Arias’ original formulation, in which reflection
coefficients obtained from x1 and x2 components might not coincidence with each other.
For the inverse problem, we present an inversion procedure to reconstruct the geometric properties
of canyon-shaped flaw by usage of reflection coefficients of Rayleigh wave at a series of frequencies.
From scattering theory for elastic waves, the scattered Rayleigh wave field is expressed as a boundary
integral equation over the flawed area. By introduction of far-field form of half-space Green’s function
and the Born approximation, it can be proven that the objective function of flaw depth is reconstructed
by inverse Fourier transform of reflection coefficients. This method gives a direct and simple inversion
approach for reconstructing locations and shapes of surface flaws, which does not require baseline
data and iteration.
For numerical verifications, the reflection coefficients obtained from forward analysis are used as input
data in the inversion procedure. It is shown that the inversion method is effective for both single and
multiple flaw cases. The method can reconstruct images especially accurately for shallow and flat notches;
while for deeper flaws, the method can still represent their geometric properties (e.g., the location, depth
and width), and can produce reliable “initial images” for further usage.
Acknowledgments: This work was supported by the National Natural Science Foundation of China (Nos.
51405225, 11502108), the Program for New Century Excellent Talents in Universities (No. NCET-12-0625),
the Natural Science Foundation of Jiangsu Province (Nos. BK20140808, BK20140037), the Fundamental Research
Funds for Central Universities (No. NE2013101), and a project Funded by the Priority Academic Program
Development of Jiangsu Higher Education Institutions (PAPD).
Author Contributions: Bin Wang and Zhenghua Qian conducted theoretical analysis; Yihui Da performed
numerical verifications; Bin Wang wrote the paper.
References
1. Graff, K.F. Wave Motion in Elastic Solids, 1st ed.; Oxford Univerisity Press: London, UK, 1975; pp. 311–393.
ISBN 0-486-66745-6.
2. Achenbach, J.D. Wave Propagation in Elasic Solids; North-Holland: Amsterdam, The Netherlands, 1975;
pp. 165–201. ISBN 0-7204-0325-1.
3. Ayter, S.; Auld, B.A. Characterization of surface wave scattering by surface breaking cracks. In Proceedings of
the ARPA/AFML Review of Progress in Quantitative NDE; Thompson, D.O., Ed.; Iowa State University Press:
Iowa City, IA, USA, 1979; pp. 498–504.
4. Achenbach, J.D. Acoustic emission from a surface-breaking crack under cyclic loading. Acta Mech. 2008, 195,
61–68. [CrossRef]
5. Lee, V.W.; Liu, W.-Y. Two-dimensional scattering and diffraction of p- and sv-waves around a semi-circular
canyon in an elastic half-space: An analytic solution via a stress-free wave function. Soil Dyn. Earthq. Eng.
2014, 63, 110–119. [CrossRef]
6. Jian, X.; Dixon, S.; Guo, N.; Edwards, R. Rayleigh wave interaction with surface-breaking cracks. J. Appl. Phys.
2007, 101, 064906. [CrossRef]
7. Zerwer, A.; Polak, M.A.; Santamarina, J.C. Rayleigh wave propagation for the detection of near surface
discontinuities: Finite element modeling. J. Nondestruct. Eval. 2003, 22, 39–52. [CrossRef]
283
Appl. Sci. 2018, 8, 427
8. Dai, Y.; Xu, B.Q.; Luo, Y.; Li, H.; Xu, G.D. Finite element modeling of the interaction of laser-generated
ultrasound with a surface-breaking notch in an elastic plate. Opt. Laser Technol. 2010, 42, 693–697. [CrossRef]
9. Wong, H.L. Effect of surface topography on the diffraction of P, SV and Rayleigh waves. Bull. Seismol. Soc. Am.
1982, 72, 1167–1183.
10. Kawase, H. Time-domain response of a semi-circular canyon for incident SV, P and Rayleigh waves calculated
by the discrete wavenumber boundary element method. Bull. Seismol. Soc. Am. 1988, 78, 1415–1437.
11. Sanchez-Sesma, F.J.; Ramos-Martinez, J.; Campillo, M. Seismic response of alluvial valleys for incident p,
s and Rayleigh waves: A boundary integral formulation. In Proceedings of the 10th World Conference of
Earthquake Engineering, Lotterdam, The Netherland, 19–24 July 1992; Ranguelov, B., Housner, G., Eds.;
Balkema: Lotterdam, The Netherlands; pp. 19–24.
12. Ávila-Carrera, R.; Rodríguez-Castellanos, A.; Sánchez-Sesma, F.J.; Ortiz-Alemán, C. Rayleigh-wave scattering
by shallow cracks using the indirect boundary element method. J. Geophys. Eng. 2009, 6, 221–230. [CrossRef]
13. Treyssède, F. Three-dimensional modeling of elastic guided waves excited by arbitrary sources in viscoelastic
multilayered plates. Wave Motion 2015, 52, 33–53. [CrossRef]
14. Zhao, X.G.; Rose, J.L. Boundary element modeling for defect characterization potential in a wave guide.
Int. J. Solids Struct. 2003, 40, 2645–2658. [CrossRef]
15. Cho, Y.; Rose, J.L. An elastodynamic hybrid boundary element study for elastic guided wave interactions
with a surface breaking defect. Int. J. Solids Struct. 2000, 37, 4103–4124. [CrossRef]
16. Kamalian, M.; Gatmiri, B.; Bidar, A.S. On time-domain two-dimensional site response analysis of topographic
structures by BEM. J. Seismol. Earthq. Eng. 2003, 5, 35–45.
17. Haghighat, A.E. Diffraction of Rayleigh wave by simple surface irregularity using boundary element method.
J. Geophys. Eng. 2015, 12, 365–375. [CrossRef]
18. Arias, I.; Achenbach, J.D. Rayleigh wave correction for the bem analysis of two-dimensional elastodynamic
problems in a half-space. Int. J. Numer. Methods Eng. 2004, 60, 2131–2146. [CrossRef]
19. Tosecký, A.; Koleková, Y.; Schmid, G.; Kalinchuk, V. Three-dimensional transient half-space dynamics using
the dual reciprocity boundary element method. Eng. Anal. Bound. Elem. 2008, 32, 597–618. [CrossRef]
20. Cook, D.A.; Berthelot, Y.H. Detection of small surface-breaking fatigue cracks in steel using scattering of
Rayleigh waves. NDT E Int. 2001, 34, 483–492. [CrossRef]
21. Masserey, B.; Fromme, P. Surface defect detection in stiffened plate structures using Rayleigh-like waves.
22. Lee, F.W.; Chai, H.K.; Lim, K.S. Assessment of reinforced concrete surface breaking crack using Rayleigh
wave measurement. Sensors 2016, 16, 337. [CrossRef] [PubMed]
23. Thring, C.B.; Fan, Y.; Edwards, R.S. Focused Rayleigh wave EMAT for characterisation of surface-breaking
defects. NDT E Int. 2016, 81, 20–27. [CrossRef]
24. Lee, F.W.; Chai, H.K.; Lim, K.S. Characterizing concrete surface notch using Rayleigh wave phase velocity
and wavelet parametric analyses. Constr. Build. Mater. 2017, 136, 627–642. [CrossRef]
25. Wang, L.; Xu, Y.; Xia, J.; Luo, Y. Effect of near-surface topography on high-frequency Rayleigh-wave
propagation. J. Appl. Geophys. 2015, 116, 93–103. [CrossRef]
26. Xia, J.; Miller, R.D.; Park, C.B. Estimation of near-surface shear-wave velocity by inversion of Rayleigh waves.
Geophysics 1999, 64, 691–700. [CrossRef]
27. Xia, J.; Nyquist, J.E.; Xu, Y.; Roth, M.J.S.; Miller, R.D. Feasibility of detecting near-surface feature with
Rayleigh-wave diffraction. J. Appl. Geophys. 2007, 62, 244–253. [CrossRef]
28. Shao, G.-Z.; Tsoflias, G.P.; Li, C.-J. Detection of near-surface cavities by generalized s-transform of Rayleigh
waves. J. Appl. Geophys. 2016, 129, 53–65. [CrossRef]
29. Koehn, D.; Meier, T.; Fehr, M.; De Nil, D.; Auras, M. Application of 2D elastic Rayleigh waveform inversion
to ultrasonic laboratory and field data. Near Surf. Geophys. 2016, 14, 461–476. [CrossRef]
30. Chen, Y.; Man, C.-S.; Tanuma, K.; Kube, C.M. Monitoring near-surface depth profile of residual stress in
weakly anisotropic media by Rayleigh-wave dispersion. Wave Motion 2018, 77, 119–138. [CrossRef]
31. Wang, B.; Hirose, S. Inverse problem for shape reconstruction of plate-thinning by guided SH-waves.
Mater. Trans. 2012, 53, 1782–1789. [CrossRef]
32. Wang, B.; Qian, Z.; Hirose, S. Inverse shape reconstruction of inner cavities using guided SH-waves in a plate.
Shock Vib. 2015, 2015. [CrossRef]
284
Appl. Sci. 2018, 8, 427
33. Wang, B.; Hirose, S. Shape reconstruction of plate thinning using reflection coefficients of ultrasonic Lamb
waves: A numerical approach. ISIJ Int. 2012, 52, 1320–1327. [CrossRef]
34. Schmerr, L.W., Jr. Fundamentals of Ultrasonic Nondestructive Evaluation; Plenum Publishing Corporation:
New York, NY, USA, 1998.
35. Achenbach, J.D. Reciprocity in Elastodynamics; Cambridge University Press: Cambridge, UK, 2003;
ISBN 052181734X.
36. Snieder, R. General theory of elastic wave scattering. In Scattering: Scattering and Inverse Scattering in Pure
and Applied Science; Pike, R., Sabatier, P., Eds.; Academic Press: Cambridge, MA, USA, 2002; pp. 528–542.
37. Kobayashi, S. Elastodynamics. In Boundary Element Methods in Mechanics, Volume 3 in Computational Methods
in Mechanics; Beskos, D.E., Ed.; North-Holland: Amsterdam, The Netherlands, 1987.
285
applied
sciences
Article
The Study of Non-Detection Zones in Conventional
Long-Distance Ultrasonic Guided Wave Inspection on
Square Steel Bars
Lei Zhang 1,2 , Yuan Yang 1, *, Xiaoyuan Wei 1 and Wenqing Yao 1
1 School of Automation and Information Engineering, Xi’an University of Technology, Xi’an 710048, China;
leizhang830102@stu.xaut.edu.cn (L.Z.); wxy@stu.xaut.edu.cn (X.W.); yaowenqing@stu.xaut.edu.cn (W.Y.)
2 School of Physics and Optoelectronic Technology, Baoji University of Arts and Sciences, Baoji 721016, China
* Correspondence: yangyuan@xaut.edu.cn; Tel.: +86-029-8231-2087
Received: 15 December 2017; Accepted: 15 January 2018; Published: 17 January 2018
Abstract: In a low-frequency ultrasonic guided wave dual-probe flaw inspection of a square steel bar
with a finite length boundary, the flaw reflected pulse wave cannot be identified using conventional
time monitoring when the flaw is located near the reflection terminal; therefore, the conventional
ultrasonic echo method is not applicable and results in a non-detection zone. Using analysis and
simulations of ultrasonic guided waves for the inspection of a square steel bar, the reasons for the
appearance of the non-detection zone and its characteristics were analyzed and the range of the
non-detection zone was estimated. Subsequently, by extending the range of the conventional detection
time domain, the envelope of the specific reflected pulse signal was extracted by a combination of
simulations and related envelope calculations to solve the problem of the non-detection zone in
conventional inspection methods. A comparison between the simulation and the experimental results
demonstrate that the solution is feasible. This study has certain practical significance for ultrasonic
guided wave structural monitoring.
Keywords: ultrasonic guided wave; nondestructive testing; square steel bar; non-detection zone
1. Introduction
Currently, ultrasonic guided wave inspection is an effective long-distance nondestructive
inspection technique [1–3]. Low-frequency ultrasonic guided waves are widely used for long-distance
ultrasonic guided wave inspection [4]. The ultrasonic pulse width is wider at low frequency than
at high frequency. Due to the influence of the pulse dispersion for long-distance transmissions,
the reflected signal pulse is wider than in short-distance transmission [5]. The superposition state
occurs easily during actual waveform measurements when the flaw is close to the terminal. In addition,
the signal trailing interference aggravates the superposition state and the flaw reflected pulse is more
difficult to distinguish. In the low-frequency ultrasonic guided wave flaw inspection experiments for a
steel square bar as described in this study, we find that the reflected pulse wave of the flaw ‘disappears’
before the first terminal reflected pulse wave (in the conventional monitoring interval) when the flaw is
near the reflection terminal; this results in a loss of the flaw’s location information, which is referred to
as the non-detection zone in this paper. Therefore, the conventional ultrasonic echo method is invalid
under this condition. The literature does not provide clear details on low-frequency ultrasonic guided
wave excitation pulse conditions and no studies have focused on providing a solution for the problem
of the non-detection zone mentioned in this study. However, the non-detection zone cannot be ignored
in studies on nondestructive testing (NDT). In this study, we designed experiments and simulations
for the inspection of a square steel bar with the focus on the above-mentioned problem.

Appl. Sci. 2018, 8, 129
Research on ultrasonic reflected pulse waves should be concerned with solving the problem of the
non-detection zone to facilitate the actual detection of flaws. To determine the ultrasonic propagation
characteristics caused by different types of flaws or structural damages, the reflected pulse wave is
used to analyze the reflection and transmission. In reference [5] the flaws of different opening angles
were analyzed and in reference [6], the first reflected pulses in a 20–45 Hz range for four sizes and three
orientations of transverse rail head flaws were investigated. Moreover, the reflected pulse wave was
used to locate the flaws. For example, the flaws in a steel rail flange were detected in reference [7] and
the orientation and position of small flaws in pipes were detected in reference [8]. In previous studies,
researchers have provided considerable advances in long-distance non-destructive testing (NDT).
However, in most studies, the flaw-reflected pulse before the first terminal reflected pulse wave was
used to locate and identify the flaw, a method we refer to a “monitoring Interval I” hereafter. Different
from the methods used in existing studies, the analysis time domain of the reflected pulse waves is
expanded to the “monitoring Interval II” in the following sections. In order to resolve the non-detection
zone problem in long-distance low-frequency ultrasonic guided wave detection, a feasible solution to
locate the flaw is proposed in this study.
To describe the non-detection zone clearly, the cause of the existence of the non-detection zone is
clarified by using a simulation with an equivalent model of low-frequency ultrasonic guided waves
for the inspection of a square steel bar in Section 2.The experimental platform and the results are
described in Section 3. The characteristics of the non-detection zone are analyzed and discussed. Based
on the characteristics of the reflected pulse waveform in the non-detection zone, a solution is proposed.
Finally, the experimental verification is described in Section 4.
2. Theory of the Non-Detection Zone

Based on previous research in low-frequency ultrasonic guided wave detection [9,10], we use
a square steel bar as an example and analyze the objective reasons for the appearance of the
non-detection zone with a finite-length boundary. Firstly, this problem associated with the ultrasonic
guided wave (UGW) becomes equivalent to the reflection of waves with different modes in a
plate [11] and is the 2D equivalent of an ultrasonic guided wave simulation model. By taking full
advantage of the characteristics of the low-frequency pulse wave [12], an analysis is performed on the
changes in the pulse signal of the flaw identification and the creation of the non-detection zone in a
conventional inspection.
Prior to the analysis of the ultrasonic guided wave echo experiments and simulations,
the following constraints are defined:
• A single low-frequency incident pulse is used in the experiments and simulation and the
waveforms of the other frequencies are not induced by the excited incident pulse at the
corresponding frequencies.
• The mode of the incident pulse wave is a single S0 mode and the frequency of the incident
pulse is lower than the S1 and A1 cut-off frequencies so that the mode is simplified and easy to
analyze [13].
• In a symmetrical plate-like structure, the reflected pulse is also a single mode S0; in an
asymmetrical plate-like structure (such as man-made flaws), the reflected pulses in the A0 mode
and S0 mode are excited [14].
• When the frequency-thickness product ( f · d) is maintained, the group velocity of the S0 and A0
modes remains unchanged under ideal conditions [15].
2.1. The Equivalent Model of Low-Frequency Ultrasonic Guided Wave in a Square Steel Bar
The square steel bar has a symmetric structure and we only focus on the propagation
characteristics of the ultrasonic guided wave propagated between the front surface and the back
surface, as shown in Figure 1a. We ignore the shear (SH) wave and the other two surfaces (the up
288
Appl. Sci. 2018, 8, 129
surface and the down surface); therefore, this is equivalent to a two-dimensional (2D) model where the
ultrasonic Lamb waves propagate along a plate [1,3].
Figure 1. The equivalent model in low-frequency (a) The equivalent model of low-frequency ultrasonic
guided wave in a square steel bar; (b) Low-frequency phase velocities of S0 and A0 modes provided by
the beam, plate and elastic layer models.
2.1.1. The Mode Selection of Ultrasonic Guided Wave

Based on previous research findings, the reasons for the mode selection of ultrasonic guided wave
in this study are as follows:
Firstly, the use of low-frequency ultrasound in long-distance ultrasonic guided wave inspection is
preferred because the attenuation of the high-frequency ultrasound transmission over long distances
is too fast, which limits the measuring distance [4]. High-frequency ultrasound waves tend to be
concentrated on the surface or in localized areas of the propagation medium [1]; therefore, the meaning
of detecting the ultrasonic guided wave as the body wave in the entire medium is lost. Because of
the multi-mode caused by the high-frequency ultrasonic excitation, the identification of detection
becomes difficult. In short, the low-frequency section and less-mode ultrasound are appropriate for
the detection requirements of long-distance ultrasonic guided wave inspection. For example, a similar
low-frequency ultrasonic guided wave equivalent simulation was applied to rails (a rod-like structure)
by Bartoli et al. [7].
Secondly, the simulated effect of an ultrasonic equivalent model is better in the low-frequency
range than in the high-frequency range. For the long wavelength in the low-frequency ultrasonic range,
the equivalent model for a rod-like structure is similar to the model of a plate-like structure for
289
Appl. Sci. 2018, 8, 129
the Lamb wave mode in a uniform medium. However, the mode distribution generated by the
high-frequency section in a rod-like structure is not similar to the Lamb wave mode, thus there is no
equivalent effect [3].
Finally, the equivalent model of low-frequency ultrasonic guided waves takes full advantage of
the proven Lamb theory and can be used to theoretically control the modes of the excited pulse wave
and the mode number [16,17]. The relationship between the S0/A0 impulsive waves and the flaw in
related research can be used to analyze the mode transformation under the finite boundary conditions
used in this study. Therefore, we use the ideal assumptions described in Section 2.
2.1.2. Theory on the Equivalent Model in Low-Frequency

The three-dimensional model of the steel square bar is set in the Cartesian coordinate system
(x, y, z), as shown in Figure 1a. The independent of z deformation is specified by the displacement
→
vector u = (u1 , u2 , 0) lying in the plane (x, y), where u1 , u2 , u3 represent the displacement variable
in the x, y, z directions respectively, and the model has a similar 2D plane-strain deformation in
the plate-structure. The similar theoretical analysis on the equivalent model for low frequency
ultrasonic wave has been proposed by Glushkov et al. [18]. We can see the similar low-frequency
phase velocities of S0 and A0 modes provided by the beam and plate models for an aluminium sample
in the reference [18], as shown in Figure 1b. In low frequency, this conclusion, that the axisymmetric
longitudinal mode in the rod with diameter d is very similar to the symmetric mode distribution in the
plate with the same thickness d, is proposed by Rose [2]. In this paper, when we focused on vibration
mode S0/A0, without regard to other vibration modes, the ultrasonic wave propagation analysis is
simplified and equivalent in low frequency.
In essence, the three-dimensional (3D) ultrasonic wave in the rod is equivalent to the
two-dimensional (2D) Lamb wave in this study, which is used as the equivalent model of the
low-frequency ultrasonic guided waves in a square steel bar. Both ultrasonic guided wave models
are based on the same acoustic propagation control equations [1]. When low frequency ultrasonic
wave excitation covers the entire incident surface, as shown in Figure 1a, the rectangular cross-section
of ABCD in the plate-like structure is used to ‘replace’ the rectangular cross section ‘A’ ‘B’ C ‘D’ in
the rod-like structure for the vibration mode analysis, and the vibration of ultrasonic wave between
the front and back surfaces in rod-like structure can be equivalent to the vibration of ultrasonic wave
between the front and back surfaces in the plate-like structure. A similar ultrasonic guided wave mode
is obtained [3], the longitudinal wave L (0, 1) and the flexural wave F (1, 1) are similar to the S0 and A0
waves respectively in a 2D plate-like ultrasonic propagation. The simulations of the pulse waveform
and the actual waveform exhibit good consistency. Compared with the 3D model, the estimation
accuracy is slightly lower but the equivalent model reduces the simulation hardware requirements
and computational costs, thereby improving the simulation efficiency.
In solid matter, as shown in Figure 1a, the vector form of the wave equation for the ultrasonic
wave can be represented as follows [1]:
→ → →
→ → λ + μ ∂u μ ∂u ∂u
(λ + μ)∇(∇• u ) + μ∇2 u + ( )∇(∇• ) + ( )∇2 =ρ (1)
ω ∂t ω ∂t ∂t
where λ and μ are the Lame constants; λ and μ denote the attenuation factor in the medium.
∂2 ∂2 ∂2
∇2 = + 2+ 2 (2)
∂x 2 ∂y ∂z
→ ∂ → ∂ → ∂
∇= i1 + i2 + i3 (3)
∂x ∂y ∂z
→
u = ( u1 , u2 , u3 ) (4)
290
Appl. Sci. 2018, 8, 129
→
The vector u is treated with the Helmholtz decomposition; u1 , u2 , u3 represent the displacement
variable in the x, y, z directions respectively; therefore, the displacement can be expressed as follows:
→ → →
u = ∇Φ + ∇ × H (5)
→
∇•H = 0 (6)
→ →
where Φ and H are the scalar and vector potentials, respectively.
Two simple wave equations can be derived as follows:
→
λ + 2μ ∂ → ∂2 Φ
(λ + 2μ + )∇2 Φ = ρ 2 (7)
ω ∂t ∂t
→
μ → ∂2 H
(λ + )∇2 H = ρ 2 (8)
ω ∂t
where ω denotes the angular frequency. If the solutions of Equations (7) and (8) are assumed to be
planar harmonic waves, then
→ → → → →
Φ( x , t) = Φ(ω )e jω ( N • x /α−t) (9)
→ → → → →
H( x , t) = H(ω )e jω ( N • x /β−t) (10)
→
where N represents the displacement vector, for the longitudinal wave, the expression of the material
parameter α is as follows:
ω
α= (11)
κre1 + i |κim1 |
where κre1 , κim1 represent the real part and the imaginary part of the plural of the longitudinal wave
number κ1 respectively.
For transverse wave, the expression of the material parameter β is as follows:
ω
β= (12)
κre2 + i |κim2 |
where κre2 , κim2 represent the real part and the imaginary part of the plural of the transverse wave
number κ2 respectively.
The stress in the z direction N3 = 0, consequently Equations (9) and (10) can degenerate into a 2D
∂
plane strain problem, then ∂z = 0.
For the longitudinal wave, the displacement vector is
→ → ∂ ∂ →
u L = ∇Φ = { , , 0} Φ (13)
∂x ∂y
The substitution of Equation (9) into Equation (13) leads to the following equations:
→ →
∂Φ( x , t) N →
u L1 = = iω 1 Φ(ω )eiω ( N1 x/α+ N2 y/α−t) (14)
∂x α
→ →
∂Φ( x , t) N →
u L2 = = iω 2 Φ(ω )eiω ( N1 x/α+ N2 y/α−t) (15)
∂y α
u L3 = 0 (16)
where u L1 , u L2 , u L3 represent compressional component in x, y, z three directions respectively.
291
Appl. Sci. 2018, 8, 129
From the formula derived in reference [3], when the longitudinal wave propagates, the normal
stress σ22 and the shear stress σ12 are expressed as:
→ →
σ22 = (α − 2N12 β2 /α)iωβeiω ( N • x /α−t) (17)
→ →
σ12 = 2iωρN1 N2 β2 /αeiω ( N • x /α−t) (18)
where ρ represents the material density, when the transverse wave propagates, the normal stress σ22
and the shear stress σ12 are expressed as:
→
σ22 = −2iωρN1 N2 βeiω ( N • x /β−t) (19)
→
σ12 = iω ( N22 − N12 ) β/αeiω ( N • x /β−t) (20)
where c p represents the phase velocity of sound, and s represents the Snell constant. The up going wave
1 1
N1 = c p s, N2 = (1 − c p 2 s2 ) 2 , and the down going wave N1 = c p s, N2 = −(1 − c p 2 s2 ) 2 , when c p = α,
longitudinal wave; when c p = β, transverse wave.
2.2. Non-Detection Zone in Simulation

To investigate the reasons for the appearance of the non-detection zone, we focus on the
simulations of the propagation of an ultrasonic guided wave in a square steel bar with a length
of 6000 mm using COMSOL Multiphysics, ver.5.0.
As shown in Figure 2, during which a 10-pulse Hanning-window function pulse signal with a
frequency of 30 kHz is used as the incident pulse.
Figure 2. The simulation settings.
The Hanning function is used as the window function:
Hanning(t) = (1 − cos(2πt/T ))/2
The type of excitation signal:
p(t) = sin(2π f 0 t) Hanning(t)
where T = 1.6 [μs].

The point (130, 30) is used as the measurement point and is the same as in the actual experiment
in Section 3. A transverse flaw with a depth of 5 mm (D0) and a width of 1 mm is created at the
292
Appl. Sci. 2018, 8, 129
displacement L with a variable value of L ∈ [3500 mm, 5500 mm] along the x direction from the
incident point and the simulation interval is set as 500 mm. (D0 is set as 5 mm for reducing the
superposition of the flaw reflected pulses). A triangular grids (λ/26) are used in this study (<λ/20
recommended minimum mesh size [19]) and the area near the location of the flaw L is refined using
a local mesh. During the simulation, the excitation energy is properly decreased so that 5 pulses are
reflected. The material parameters of the steel square bar are the following:
Young’s modulus: E = 200 [GPa]
Density: ρ = 7850 [kg/m3 ]
Poisson’s ratio: ν = 0.3
Taking advantage of the periodicity of the terminal reflection, as shown in Figure 3a, the time
domain is divided into the monitoring time Interval I, the monitoring time Interval II, the monitoring
time Interval III, and so on. The pulse waves of the multiple reflections are labeled as the incident
pulse wave, the first terminal reflection pulse wave, the second terminal reflection pulse wave, and so on.
Figure 3. Cont.
293
Appl. Sci. 2018, 8, 129
Figure 3. Waveforms of the reflected pulses in the simulation, (a) D0 = 0 mm; (b) D0 = 5 mm,
L = 3500 mm; (c) D0 = 5 mm, L = 4000 mm; (d) D0 = 5 mm, L = 4500 mm; (e) D0 = 5 mm, L = 5000 mm;
(f) D0 = 5 mm, L = 5500 mm.
As shown in Figure 2b, at L = 3500 mm, an S0 mode wave is emitted from the incident face and
is then reflected by the transverse flaw [20]; then, after the reflection, two different types of waves,
S0 and A0, are generated. In Interval I, the peak points of the S0 mode and A0 mode flaw-reflected
pulse waves are (0.001598, 0.1206) and (0.00205, −0.1729), respectively. Based on the simulation results,
the S0 mode flaw-reflected pulse wave is calculated to verify whether the detected location of the flaw
was accurate.
It is known that the first terminal-reflected pulse wave and the initial pulse exhibit the maximum
amplitude at ((0.0002036, 1) and (0.002642, 1.195)); the S0 wave group velocity (CgS0 ) was calculated as:
CgS0 = Δs/Δt = (6 − 0.13) × 2/(0.002642 − 0.0002036) = 4814.6325 [m/s] (21)
where Δs and Δt represent the propagation distance and the corresponding time respectively. Using
the calculated CgS0 , the flaw-induced S0 mode reflected wave is verified. The S0 mode pulse wave
294
Appl. Sci. 2018, 8, 129
appears at 0.001598 s and L0 denotes the distance between the transverse flaw and the initial face,
the following expression can be acquired:
L0 = (CgS0 × Δt)/2 + 0.13 = CgS0 × (0.001598 − 0.0002036)/2 + 0.13 = 3.4868 [m] (22)
The detected location is quite close to the actual position of the flaw (3.5 m). The relative deviation
σ0 can be calculated by:
σ0 = (3.5 − 3.4868)/3.5 = 0.3771% (23)
As shown in Figure 2b, the A0 mode flaw-reflected pulse wave first appears at 0.00205 s. An S0
mode ultrasonic pulse wave was excited to the transverse flaw at 3.5 m and the mode changed from S0
to A0 when passing the asymmetric flaw. Therefore, the propagation time of the A0 mode ultrasonic
pulse wave after mode transmission is equal to:
Cg A0 = Δs/Δt = (3.5 − 0.13)/(0.00205 − 3.5/CgS0 ) = 2547.2411 [m/s] (24)
The flaw at 4000 mm (Figure 3c) can also be calculated in the same way so that its location can
be determined using the reflected A0 mode and S0 mode waves. As shown in Figure 3d,e, the flaw
at 4500 mm and 5000 mm can be only located by the S0 mode reflected pulse in the conventional
inspection because the A0 mode pulse wave, which is in the superposition, is not in Interval I. As shown
in Figure 3f, the times of the S0 mode and A0 mode flaw-reflected pulse waves in Interval I when
D0 = 5 mm at 5500 mm are then calculated.
The appearance time of the S0 mode pulse wave TS0 can be calculated using:
TS0 = 0.0002026 + (5.5 − 0.13) × 2/CgS0 = 0.0024333 [s] (25)
The appearance time of the A0 mode pulse wave TA0 can be calculated using:
TA0 = 5.5/CgS0 + (5.5 − 0.13)/Cg A0 = 0.0032506 [s] (26)
This analysis shows that, when the value of L increases along the x-direction and approaches
the length of the terminal, the appearance time of the S0 mode flaw-reflected pulse wave and the A0
mode flaw-reflected pulse wave gradually approach the time of the first terminal reflection pulse wave,
even then the appearance time of the A0 mode flaw-reflected pulse appears in the monitoring time
Interval II at a given value of L. For the above simulation, at L = 5500 mm, the S0 mode flaw-reflected
pulse wave is superimposed by the first terminal reflection pulse wave and the A0 mode flaw-reflected
pulse wave is superimposed by these reflection pulse waves of the monitoring time Interval II.
In particular, the calculation (see Equation (26)) indirectly suggests that, in the monitoring
Interval II, the highest reflected pulse is not the delay of the A0 mode flaw-reflected pulse in the
conventional inspections, as shown in Figure 3f. Under ideal conditions, even without the interference
of the signal trailing, the actual S0 or A0 mode flaw reflection pulses cannot be easily identified and the
conventional detection method based on a single mode flaw-reflected pulse wave is invalid. For actual
measurements, the trailing phenomenon of the pulse wave has to be considered. This is the state
of the non-detection zone referred to in this study. It is known that the information of the flaw is
superimposed by the first terminal reflection pulse wave and various types of reflection pulse waves
in the monitoring time Interval II described in this section. In the following section, we observe the
actual waveform of the ultrasonic pulse wave under the condition of the non-detection zone in an
actual experiment.
3. Experimental Section
In this section, the experimental platform and the experimental results are presented and the
occurrence of the non-detection zone in square steel inspections is briefly described.
295
Appl. Sci. 2018, 8, 129
As shown in Figure 4a, a 30# (30 mm × 30 mm) square steel bar with a length of 6000 mm is used
in this study; the length is sufficient to weaken the effect induced by signal trailing and facilitates the
signal identification. These customized ultrasonic longitudinal-wave probes [21,22] with a central
frequency of 30 kHz are installed using the dual-probe echo method. The signals are acquired and
displayed using an oscilloscope (Tektronix MD04104) and the waves are recorded and the data are
captured using a wave recorder (HIOKI MR8875-30), as shown in Figure 4b.
Figure 4. Establishment of the experimental platform (a) square steel bar with a length of 6000 mm;
(b) picture of experimental instruments; (c) the diagram of the establishment of experimental platform.
The Hamming pulse signal (10 cycles at the center frequency of f 0 = 30 [kHz]) is set by the signal
source, which is loaded to the ultrasonic probe via a power amplifier. Transverse flaws with a width
of 1 mm and a depth of 0~20 mm (D0) are created at a displacement L along the x-direction from the
incident point. Because the thickness of the steel square bar is 30 mm and the length of the actual
ultrasonic probe is 10 mm, a point (130.30) is selected as the measuring point. The excitation voltage
amplitude is adjusted to acquire five reflected pulses and the waveforms of the reflected pulses are
monitored in the time domain. Figure 4c shows the overall layout of the experimental platform.
In a dual-probe echo long-distance inspection with the customized longitudinal-wave probe used
in the experiment, the acquired reflected pulse wave of the transverse flaw occurs before reaching
the first terminal; subsequently, the flaw can be located based on the modal wave velocity. However,
both the experimental and the simulation results of the square steel inspection indicate that no reflected
pulse is identified before the first terminal reflection pulse wave when the flaw is at a certain distance
(L ∈ (5300 mm, 6000 mm) from the excitation point. The experimental result at L = 5500 mm is used as
an example (Figure 5). In the conventional monitoring Interval I, the reflected pulse wave with the flaw
cannot be identified. According to the experimental data, even when the depth of the flaw (D0) reaches
approximately 20 mm (i.e., the flaw represents almost 75% of the 30-mm thickness), no echo information
296
Appl. Sci. 2018, 8, 129
is detected in the conventional monitoring Interval I. In actual measurements, when the ultrasonic
pulse excitation ends, there are attenuation oscillations in the signal, namely the trailing phenomenon
of the pulse wave. Many factors affect signal trailing, such as the manufacturing technique of the
probes, the capacitance shock effect of the excitation circuit, and the sound absorption effect of the
substrate. The trailing phenomenon is common in long-distance inspections with low-frequency
ultrasonic guided waves [23,24]. In addition, the signal trailing compounds the difficulty of identifying
the flaw-reflected pulse wave in the non-detection zone.
Figure 5. The reflected pulse waves with different values of D0 at 5500 mm.
At L = 5500 mm, the S0 mode and A0 mode flaw-reflected pulse waves are superposed by the
first terminal-reflected pulse wave, various types of reflection pulse waves in the monitoring time
interval II, and the pulse signal trailing. Therefore, in a conventional inspection, the non-detection
zone can be easily determined in the actual experiment. A similar experimental phenomenon can also
occur in the ultrasonic flaw detection of long rod-like structures and cannot be ignored for NDT.
297
Appl. Sci. 2018, 8, 129
4. Analysis and Discussion

In this section, we analyze the range of the non-detection zone and the propagation characteristics
of the ultrasonic guided wave in the non-detection zone using a simulation and experiments.
Subsequently, we propose a feasible solution for the practical problems.
4.1. The Range of the Non-Detection Zone

The range of the non-detection zone is a major concern in non-destructive inspections. Based
on the analysis in Section 2.2, we used a dual-probe echo method for the location of the flaws and a
low-frequency of the incident pulse. If the flaw is a certain distance from the terminal, the S0 mode
flaw-reflected pulse is superposed by the first terminal-reflected pulse wave, while the A0 mode
flaw-reflected pulse is superposed by the multiple reflected pulse wave and the pulse signal trailing in
the monitoring interval II; as a result, the non-detection zone occurs. Therefore, the superposition time
point of the S0 mode flaw-reflected pulse and the first terminal-reflected pulse wave can be considered
as given the occurring conditions. Figure 6 shows the first terminal-reflected pulse wave and the
S0 mode flaw-reflected wave.
Figure 6. Schematic of the distance covered by the S0 mode pulse wave in (a) the first terminal-reflected
pulse wave and (b) the flaw-reflected pulse wave in Interval I.
Based on these occurring conditions, the range of the non-detection zone can be described as:
CgS0 × (Δh/2)/2 ≥ ΔL > 0 (27)
where CgS0 is related to the frequency-thickness product ( f · d) of the incident pulse wave,
ΔL denotes the distance between the flaw and the terminal, Δh denotes the pulse width of the
first terminal-reflected pulse wave (Figures 3a and 5), which is connected with the incident pulse
wave and the self-characteristic of the customized longitudinal-wave probe, and ΔL denotes the
distance between the flaw and the terminal reflection boundary. It should be noted that low-frequency
ultrasound is normally used as the core frequency of the excitation pulse in long-distance ultrasonic
guided wave inspections [24]. This has many advantages such as reducing the excitation mode number
and extending the range of the long-distance detection [25]. However, the disadvantages cannot
be ignored. One disadvantage is that the ultrasonic pulse width (Δh) is wider at a low frequency
298
Appl. Sci. 2018, 8, 129
than at a high frequency. In addition, in a low-frequency ultrasonic guided wave propagated over a
long distance [26], the width of the first terminal reflected pulse wave (Δh) increases because of the
dispersion effect, thereby expanding the range of the non-detection zone.
The first terminal reflection pulse wave is the result of the overlap of the reflected pulse waves
from the end surface and the incidence surface. The width of the first terminal reflection pulse wave
(Δh) varies with the transmission distance. This is related to the measuring points, the detection
distance, the characteristic of the probe, and so on. In this study, the range of the non-detection zone
for the inspection of the 30# square steel bar with a length of 6000 mm is used as an example; the width
of the first reflected pulse (Δh) is 0.000822 s in the actual experiment (Figure 5) and 0.0005371 s in the
simulation (Figure 3a).
The governing equation for the Rayleigh-Lamb waves in a plate-like structure [1] for the symmetric
mode is:
2
tan(qd)/ tan( pd) = −4k2 pq/(q2 − k2 ) (28)
and for the anti-symmetric mode:
2
tan(qd)/ tan( pd) = −(q2 − k2 ) /4k2 pq (29)
where p2 = (ω/c L ) − k2 , q2 = (ω/c T ) − k2 , the wave number k = ω/c P , the longitudinal wave
velocity c L , the shear velocity c T , the phase velocity c P = (ω/2π )λ, the group velocity Cg and the
thickness of the plate is d.
The formula of the group velocity theory is:
Cg = dω/dk (30)
By using (Equations (28)–(30)) and selecting the material parameters for the iron plate and
the curve of the S0 mode is calculated. According to the theoretical S0 wave group velocity
(CgS0 ) in the dispersion curve [27] and the width of the first terminal-reflected pulse (Δh) in the
experiment, the curves of the non-detection zone range are calculated and are shown in Figure 7.
For the ferrous material used in this study, for which the frequency-thickness product ( f · d) is
30 × 30 kHz·mm, the range of the corresponding non-detection zone (ΔL) is 0.6390 m in the 6000 mm
long steel square bar; This is basically coincident with the simulations. During the entire process,
below a value of 2.5 MHz·mm, the range of the non-detection zone decreases with the increase
in the frequency-thickness product; when the frequency-thickness product is about 2.5 MHz·mm,
the minimum length of the non-detection zone is 0.2 m. In the range of 2.5 MHz·mm to 5.0 MHz·mm,
the range of the non-detection zone increases with the increase in the frequency-thickness product.
When the ultrasonic guided wave has a frequency-thickness product of 5.0 MHz·mm, the range of
the non-detection zone varies slightly. In an actual long-distance ultrasonic guided wave propagation
process, the range of the non-detection zone expands with the widening of the first terminal-reflected
pulse wave, which is a new reference factor to be considered. Similarly, the simulations are conducted
using different materials [1] and the other conditions remain unchanged, as the simulated non-detection
zone shows in Figure 7. The materials have different properties, resulting in the increase in the
minimum length of the non-detection zone. Because the minimization of the non-detection zone is
desirable, alloying may present a solution in the low-frequency range.
Although the flaw’s pulse signal is lost in monitoring Interval I, the reflected pulses carrying the
flaw information can still be detected in other monitoring intervals, thus allowing the determination of
the flaw’s location. By analyzing the simulation results that satisfy the conditions for the appearance
of the non-detection zone, it can be found that the overall pulse waveform exhibits apparent variations
in monitoring Interval II.
299
Appl. Sci. 2018, 8, 129
Figure 7. Curves of the non-detection zone ranges for different materials.
4.2. Characteristics of the Reflected Pulse Wave in Monitoring Interval II

This study attempts to locate the flaw in the non-detection zone using the reflected pulse wave
in monitoring Interval I and monitoring Interval II based on a simulation and experimental results.
The primary reasons are listed below. Firstly, the amplitude of the reflected pulse in the monitoring
Interval II exceeds that encountered in the following monitoring intervals. Secondly, because the
existence of the flaw, an increasing number of derivative pulse waves will occur as the propagation
time increases; the condition of the pulse wave superposition is simpler in monitoring Interval II
than in the following monitoring Intervals. Therefore, we used the square steel bar with a length of
6000 mm as an example and analyzed the variation of the reflected pulse waveform in monitoring
Interval II.
Due to the variation of the pulse width of the terminal-reflected pulse wave and the waveform
superposition in the actual experiments, the non-detection zone is observed in advance when
L = 5300 mm. Therefore, this study focused on the simulation range of [5300 mm, 5900 mm] with
D0 = 10 mm and the simulation interval of 100 mm. The analysis indicates that the amplitude changed
with the changes in D0, leading to new changes in the envelope superposition of the incident pulse in
Interval II and especially in the superposition of the amplitudes. We used D0 = 10 mm as an example
of the representative and significant waveform change; other values of D0 can also be used for the
same method.
Figure 8 shows the simulation results. It can be observed that, as the flaw location (L) changes,
the amplitude of the superposed pulse wave exhibits complex changes in the time domain, especially
in Interval II. Moreover, as the flaw location (L) changes, the superposition envelope of multiple
reflected pulses changes significantly in monitoring Interval II, i.e., both of these factors showed a
one-to-one correspondence. Within the monitoring Interval II, there exists multiple reflected wave
energy; at first, at around L = 5300~5500 mm, the highest reflected pulse gradually moves backward;
at around L = 5600~5700 mm, the two highest reflected pulses appear in adjacent time-domains;
at around L = 5800~5900 mm, the flaw position is close to the terminal, the energy gradually becomes
concentrated, and the amplitudes of these flaw-reflected pulse waves are significantly higher than the
amplitude of the first terminal reflected pulse wave for the two conditions. Therefore, we attempted
to use the above-mentioned correspondence and estimated the location of the flaw through reverse
deduction based on envelope comparison.
300
Appl. Sci. 2018, 8, 129
Figure 8. Simulated waveforms of the square steel bar with a length of 6000 mm at L ∈ [5300 mm, 5900 mm].
301
Appl. Sci. 2018, 8, 129
4.3. Approximation Based on Envelope Comparison

Although the reflected pulses show complex superposition patterns in Interval II, the propagation
speed of the fundamental mode remains relatively stable and the superposition state varies
correspondingly as the flaw location L changes along the x-direction. If the energy of the excitation
pulse signal, the flaw location L, and D0 were fixed, the propagation energy would attenuate.
Compared with the condition without any flaws, the amplitude of the pulse overall decreases but the
amplitude of the pulse reflected by the flaw increases. The difference in the envelope between these
two different conditions (with and without the flaw) of the pulse waves can be used to identify the
reflected pulse.
Because the ultrasonic reflected pulse wave from the flaws located in monitoring Interval I is
invalid, we decided to use the envelope wave of the pulse that is extended to monitoring Interval II to
locate the flaws. To resolve the interference problem caused by the trailing of the actual experimental
data and the unexpectedly generated waves, firstly, we use the envelope of the waveforms of two
areas (for example, D0 = 0 mm and D0 = 10 mm) in the simulation and in the experiment and calculate
the difference in the envelope amplitude of the two areas; these are used as basic data for comparison.
Firstly, by observing these envelopes shown in Figure 9a–f, a coarse division of the flaw areas is
proposed based on the analysis in Section 4.2.
• Situation 1: in the actual experimental envelope during monitoring Interval II, the reflected pulse
with only one envelope amplitude that is larger than the amplitude of the terminal reflected pulse
wave is categorized into a zone ranging from L = 5800 mm to L = 5900 mm;
• Situation 2: the reflected pulses with two similar highest envelope amplitudes are categorized
into a zone ranging from L = 5600 mm to L = 5700 mm;
• Situation 3: the remaining flaw detection pulses with different amplitudes are categorized into a
zone from L = 5400 mm to L = 5500 mm.
Secondly, we selected the effective time domain intervals for comparison (from 7.649 × 10−4 s to
4.653 × 10−3 s intervals within the monitoring Intervals I and II) to avoid the interference caused by
the trailing of the incident pulse and the second terminal-reflected pulse wave, as shown in Figure 9f.
Thirdly, the envelope theory (ET) is known as an auxiliary field method [28,29]; therefore, we could
use this convenient method for the wave comparison between the simulation and the measurements.
The envelope correlation algorithm is applied to compute the maximum correlation coefficient in order
to choose the most similar envelope waveforms in a quantitative approach and locate the flaws.
The discretization cross-correlation function can be represented as follows:
N −1
Rse (k ) = ∑ s(n)e(n − k ) k = 0, 1, 2, · · · , M − 1 (31)
n =0
where Rse (k ) denotes the correlation function for the simulation data s(n) and the experimental data
e(n). N is the cumulative average number and k is the time-lapse sequence number, M is a positive
integer. The formula for the envelope correlation is expressed as follows:
#
$ N −1
$ N −1
ρ = Rse (0)/% ∑ s2 (n) ∑ e2 (n) (32)
n =0 n =0
The scan comparisons are not required (k = 0) in Equation (31). The envelope correlation
coefficient (ρ) of the experimental data and the simulation data are obtained using Equation (32),
and the envelope whose simulation data is closest to the experimental data envelope is selected as the
indicator for locating the corresponding flaw.
302
Appl. Sci. 2018, 8, 129
Figure 9. Cont.
303
Appl. Sci. 2018, 8, 129
Figure 9. Envelope calculation results of the pulse wave at different locations of flaw: (a) L = 5400 mm
D0 = 10 mm (b) L = 5500 mm D0 = 10 mm (c) L = 5600 mm D0 = 10 mm (d) L = 5700 mm D0 = 10 mm
(e) L = 5800 mm D0 = 10 mm (f) L = 5900 mm D0 = 10 mm.
304
Appl. Sci. 2018, 8, 129
From the above analysis, the approximation is based on the inverse comparison with the
simulation envelope in the non-detection zone. As shown in Figure 10, the main steps are as follows:
Data pre-processing: we prepared the envelope data in the experiment and the simulation without
the flaw. The experimental data can be replaced by the experimental data of a similar square steel bar
without flaws in practical measurements. For the simulation data, we can use D0 = 0 mm.
• Step 1: The experimental data with the flaw obtained in the non-detection zone are processed
using the envelope calculation and we subtract the envelope of the experimental data without the
flaw to reduce the effect of the signal trailing. Then the result of the envelope is extracted in the
effective comparison interval (from 7.649 × 10−4 s to 4.653 × 10−3 s).
• Step 2: Based on the obtained experimental envelope (Envelopment 2–Envelopment 1),
the simulation range (L ∈ [L1 , L2 ]) is selected according to the rough classification described
in Section 4.3 to reduce the computation cost.
• Step 3: According to the comparison requirements, the variable value of L is selected in the
appropriate range (L1 mm ≤ L1 + kΔx ≤ L2 mm, Δx = 100 mm, k ∈ N), the depth of the flaw (y)
in the experiment is selected as the value of D0 in the simulation (0 mm ≤ D0 = y < 30 mm),
and the simulation results have been processed using the envelope calculations respectively.
For every simulation envelope, the envelope of the simulation without the flaw (D0 = 0 mm) is
subtracted. Then the results of these envelopes are extracted in the comparison interval as a series
of comparisons.
• Step 4: The correlation coefficients between the experimental envelope data from Step 1 and
every simulation envelope data from Step 3 are determined and the result returns the maximum
correlation coefficient ρ; subsequently, the estimated position of the flaw (L) is determined
according to the reverse correspondence.
Final data processing: By changing the value of Δx, a similar approximation is used near the
initial estimated value of L to approximate the real value of L in a further simulation.
Figure 10. The schematic diagram of the envelope approximation method.
305
Appl. Sci. 2018, 8, 129
4.4. Experimental Validation

In order to localize the flaw in the non-detection zone, the different envelopes of the simulated
envelope waveforms (Envelopment 2–Envelopment 1) are used as the evaluation standard to compare
with those of the experimental data. This section describes the validation of the experiment. In order
to depict the validation results, instead of strictly following the steps mentioned in Section 4.3,
the preliminary classification in Step 2 was omitted. The variable value of L is selected from an
appropriate range (L1 = 5400 mm, L2 = 5900 mm) in the simulation. D0 = 10 mm is used as an example
of the experimental data to determine the location of the flaw (L).
Figure 11a shows the experimental pulse waveform of the square steel bar without the flaw
(D0 = 0 mm) and the experimental and simulated pulse waveforms of the square steel bar with the
flaw (D0 = 10 mm). Using the envelope calculation, it can be observed that the envelope waveform in
monitoring Interval II reduces the interference of the signal trailing in the experiments. Compared
with the simulation data, it is evident that the peak points of the difference envelopes in the experiment
are consistent, as shown in the red dotted line. The time relative deviation σ1 can be calculated by:
σ1 = (0.003187 − 0.003107)/0.003187 = 2.5101% (33)
The flaw’s reflected information that is hidden and the superposition envelope of the flaw-reflected
pulse in the trailing are extracted (Envelopment 2–Envelopment 1).
Figure 11. Location of the flaw through the envelope approximation method: (a) The envelope
of experimental data processing (b) The comparison of multiple envelopes in the effective
comparison interval.
306
Appl. Sci. 2018, 8, 129
Subsequently, the results of the different envelopes (Envelopment 2–Envelopment 1) in the

simulation and the experiment are extracted in the comparison interval, as shown in Figure 11b.
The correlation coefficients (ρ) of the experimental data and the different simulation data are
determined. Because the simulated envelope waveform is used as the evaluation standard, the noise
introduced by the real application cannot be eliminated by the correlation algorithm; therefore,
the correlation coefficient is relatively low. However, we can still determine that the correlation
coefficient, which is negative, should be discarded in the range from L = 5800 mm to L = 5900 mm;
in the range from L = 5600 mm to L = 5700 mm, the envelopes of the waveform are not well matched.
However, in the range from L = 5400 mm to L = 5500 mm, the approximated simulation waveforms in
the flaw area (L = 5500 mm) have a significant correlation coefficient of 0.3430; therefore, L = 5500 mm
is selected as the estimated value of the location of the flaw. If these calculating steps are strictly carried
out according to the solution in the previous section, based on the preliminary rough classification, it is
found that the actual measurement data range from L = 5400 mm to L = 5500 mm and we can obtain
similar results rapidly to approximate the real value of L in further simulations.
Therefore, the flaw in the non-detection zone can be identified using the inverse correspondence.
Using an equivalent mode and approximation algorithms, this methodology provides estimations of
the locations of flaws relatively accurately.
Based on the analysis in Section 2.1.1, when the equivalent plate thickness (d) is constant, as long
as the frequency is below the cutoff frequency of A1 and S1, the excitation frequency in this range can
be used for low-frequency ultrasound mode selection. For example, the proposed analysis is applicable
to the ultrasonic probe core frequency range of 20–60 kHz low-frequency ultrasonic guided wave
detection. The concept of detecting flaws using simulations also applies to different types of flaws.
In this study, we extended the analysis to a triangular flaw. This approach also resulted in a similar
positioning effect. When f · d is known, the modal phase velocity and the group velocity basically
remain unchanged, whereas there is a one-to-one correspondence between the type of the reflected
pulse wave and the location of the flaw. Therefore, in this study, we use a triangular flaw (40 kHz,
D0 = 5 mm, L = 5700 mm) as another example.
Based on the data analysis shown in Figure 12, we can obtain a similar conclusion; the flaw area
(L = 5700 mm) has a significant correlation coefficient (ρ) of 0.3091. Therefore, L = 5700 mm is then
selected as the estimated value of L (the location of the flaw). The solution remains valid, but the
characteristics of the waveform superposition have changed, which can be used to identify the triangle
type of the flaw.
Figure 12. Cont.
307
Appl. Sci. 2018, 8, 129
Figure 12. Location of a triangle flaw. (a) The triangle type of the flaw on the steel square bar; (b) The
comparison of multiple envelopes in the effective comparison interval for the triangle flaw.
The relevant analysis and the solution presented in this paper are still valid for low-frequency
ultrasonic guided wave detection in a rod-like structure medium. We can also obtain a similar
conclusion for the non-detection zone problem for a rail structure. For example, a rail experiment using
60 kHz as the actual measurement frequency was conducted by Rose [9]. Although the experiment
used different excitation pulse waveforms, which were affected by other derived modes wave and
pulse trailing, the superimposed reflected pulse waveform obtained in the second monitoring time
domain and the peak value was far higher than the peak value of the first reflected pulse; this result is
very similar to the results of the characteristic analysis in this study.
5. Conclusions
Previous studies have provided no clear statements or solutions regarding the non-detection zone
in low-frequency ultrasonic guided waves. However, the existence of the non-detection zone cannot
be ignored in NDT. In this study, we located the non-detection zone in long-distance conventional
ultrasonic guided wave inspections of a square steel bar and investigated the reasons and conditions for
the appearance of the non-detection zone through simulations. Further, by analyzing the characteristics
of the simulated pulse wave when the non-detection zone appears, the range of the non-detection zone
and a solution are proposed. The simulation envelope is used for a relative comparison; by suppressing
the interference of the ultrasonic signal trailing using the envelope calculation, the flaw is located by a
comparison with the simulation results of the multi-pulse superimposed envelope in the extended
time domain (the monitoring time Intervals I and II). The simulation results are verified with actual
measurements. The results of this study provide a feasible solution for locating the non-detection
zone and serve as a powerful supplement for conventional long-distance ultrasonic guided wave
inspections. This type of non-detection zone can also occur in the ultrasonic flaw detection of long
rod-like structures; therefore, this study has certain applicability and practical significance for other
structures. However, due to the influence of actual field factors on the actual detection of the pulse
energy loss, the amplitude deviation between the equivalent simulation and the actual measurement is
large in this study and is not conducive to the prediction of the depth (D0) of the flaw. For this reason,
it is necessary to improve the approximate model of the low-frequency ultrasonic guided wave and
the estimation of various types of flaws to provide a new approach that is widely applicable.
308
Appl. Sci. 2018, 8, 129
Acknowledgments: This work was supported by the Key Research and Development Program of
Shaanxi Province (Grant No. 2017ZDXM-GY-130), Xi’an City Science and Technology Project (Grant No.
2017080CG/RC043 (XALG009)).
Author Contributions: Yuan Yang and Lei Zhang performed the theoretical analysis, conceived and designed
the experiments; Lei Zhang, Wenqing Yao and Xiaoyuan Wei performed the experiments and analyzed the data;
Lei Zhang wrote the paper.
References
1. Rose, J.L. Ultrasonic Guided Waves in Solid Media, 1st ed.; Cambridge University Press: New York, NY, USA,
2014; ISBN 978-110-70489-59.
2. Rose, J.L. Ultrasonic Waves in Solid Media, 1st ed.; Cambridge University Press: New York, NY, USA, 1999;
ISBN 0521640431.
3. Ai, C.Y.; Li, J.; Liu, Y.; Ni, T. Ultrasonic Test Theory and Technology for Multilayered Bonding Structure, 1st ed.;
National Defence Industry Press: Beijing, China, 2014; ISBN 978-711-80958-07.
4. Verma, B.; Mishra, T.K.; Balasubramaniam, K.; Rajagopal, P. Interaction of low-frequency axisymmetric
ultrasonic guided waves with bends in pipes of arbitrary bend angle and general bend radius. Ultrasonics
2014, 54, 801–808. [CrossRef] [PubMed]
5. Ryue, J.; Thompson, D.J.; White, P.R.; Thompson, D.R. Decay rates of propagating waves in railway tracks at
high frequencies. J. Sound Vib. 2009, 320, 955–976. [CrossRef]
6. Gravenkamp, H.; Prager, J.; Saputra, A.A.; Song, C. The simulation of lamb waves in a cracked plate using
the scaled boundary finite element method. J. Acoust. Soc. Am. 2012, 132, 1358–1367. [CrossRef] [PubMed]
7. Bartoli, I.; Scalea, F.L.D.; Fateh, M.; Viola, E. Modeling guided wave propagation with application to the
long-range defect detection in railroad tracks. NDT E Int. 2005, 38, 325–334. [CrossRef]
8. Campos-Castellanos, C.; Gharaibeh, Y.; Mudge, P.; Kappatos, V. In the application of long range ultrasonic
testing (LRUT) for examination of hard to access areas on railway tracks. Railw. Cond. Monit. Non-Destr. Test.
2012, 1–7. [CrossRef]
9. Rose, J.L.; Avioli, M.J.; Mudge, P.; Sanderson, R. Guided wave inspection potential of defects in rail.
10. Wu, J.; Wang, Y.; Zhang, W.; Nie, Z.; Lin, R.; Ma, H. Defect detection of pipes using Lyapunov dimension of
Duffing oscillator based on ultrasonic guided waves. Mech. Syst. Signal Proc. 2016, 82, 130–147. [CrossRef]
11. Yao, W.; Sheng, F.; Wei, X.; Zhang, L.; Yang, Y. Propagation characteristics of ultrasonic guided waves in
continuously welded rail. Mod. Phys. Lett. B 2017, 1740075. [CrossRef]
12. Su, Z.; Ye, L.; Lu, Y. Guided lamb waves for identification of damage in composite structures: A review.
J. Sound Vib. 2006, 295, 753–780. [CrossRef]
14. Mirahmadi, S.J.; Honarvar, F. Application of signal processing techniques to ultrasonic testing of plates by
lamb wave mode. NDT E Int. 2011, 44, 131–137. [CrossRef]
15. Pai, P.F.; Deng, H.; Sundaresan, M.J. Time-frequency characterization of lamb waves for material evaluation
and damage inspection of plates. Mech. Syst. Signal Proc. 2015, 62–63, 183–206. [CrossRef]
16. Deng, Q.T.; Yang, Z.C. Scattering of S0 lamb mode in plate with multiple damages. Appl. Math. Model. 2011,
35, 550–562. [CrossRef]
17. Wan, X.; Tse, P.W.; Chen, J.; Xu, G.; Zhang, Q. Second harmonic reflection and transmission from primary
s0 mode lamb wave interacting with a localized microscale damage in a plate: A numerical perspective.
18. Glushkov, E.; Glushkova, N.; Eremin, A.; Giurgiutiu, V. Low-cost simulation of guided wave propagation in
notched plate-like structures. J. Sound Vib. 2015, 352, 80–91. [CrossRef]
19. Rucka, M. Experimental and numerical study on damage detection in an l-joint using guided wave
propagation. J. Sound Vib. 2010, 329, 1760–1779. [CrossRef]
20. Le, C.E.; Castaings, M.; Hosten, B. The interaction of the s0 lamb mode with vertical cracks in an aluminium
plate. Ultrasonics 2002, 40, 187–192. [CrossRef]
309
Appl. Sci. 2018, 8, 129
21. Yang, Y.; Wei, X.; Zhang, L.; Yao, W. The effect of electrical impedance matching on the electromechanical
characteristics of sandwiched piezoelectric ultrasonic transducers. Sensors 2017, 17, 2832. [CrossRef]
[PubMed]
22. Wei, X.; Yang, Y.; Yao, W.; Zhang, L. Pspice modeling of a sandwich piezoelectric ceramic ultrasonic
transducer in longitudinal vibration. Sensors 2017, 17, 2253. [CrossRef] [PubMed]
23. Zhao, Y.; Li, F.; Cao, P.; Liu, Y.; Zhang, J.; Fu, S.; Zhang, J.; Hu, N. Generation mechanism of nonlinear
ultrasonic lamb waves in thin plates with randomly distributed micro-cracks. Ultrasonics 2017, 79, 60–67.
[CrossRef] [PubMed]
24. Loveday, P.W. Simulation of piezoelectric excitation of guided waves using waveguide finite elements.
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2008, 55, 2038. [CrossRef] [PubMed]
25. Ryue, J.; Thompson, D.J.; White, P.R.; Thompson, D.R. Investigations of propagating wave types in railway
tracks at high frequencies. J. Sound Vib. 2008, 315, 157–175. [CrossRef]
26. Zhang, Y.; Li, D.; Zhou, Z. Time reversal method for guided waves with multimode and multipath on
corrosion defect detection in wire. Appl. Sci. 2017, 7, 424. [CrossRef]
27. Zhang, H.Y.; Xu, J.; Ma, S.W.; Ta, D.A. The high-frequency scattering of the s0 lamb mode by a circular blind
hole in a plate. Phys. Procedia 2015, 70, 455–458. [CrossRef]
28. Santacruz, J.; Tardón, L.; Barbancho, I.; Barbancho, A. Spectral envelope transformation in singing voice for
advanced pitch shifting. Appl. Sci. 2016, 6, 368. [CrossRef]
29. Semay, C. The hellmann–feynman theorem, the comparison theorem, and the envelope theory. Results Phys.
2015, 5, 322–323. [CrossRef]
310
applied
sciences
Article
Research on a Rail Defect Location Method Based on
a Single Mode Extraction Algorithm
Bo Xing 1 , Zujun Yu 1,2 , Xining Xu 1,2, *, Liqiang Zhu 1,2 and Hongmei Shi 1,2
1 School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing 100044, China;
15116333@bjtu.edu.cn (B.X.); zjyu@bjtu.edu.cn (Z.Y.); lqzhu@bjtu.edu.cn (L.Z.); hmshi@bjtu.edu.cn (H.S.)
2 Key Laboratory of Vehicle Advanced Manufacturing, Measuring and Control Technology
(Beijing Jiaotong University), Ministry of Education, Beijing 100044, China
* Correspondence: xnxu@bjtu.edu.cn
Abstract: This paper proposes a rail defect location method based on a single mode extraction
algorithm (SMEA) of ultrasonic guided waves. Simulation analysis and verification were conducted.
The dispersion curves of a CHN60 rail were obtained using the semi-analytical finite element method,
and the modal data of the guided waves were determined. According to the inverse transformation
of the excitation response algorithm, modal identification under low-frequency and high-frequency
excitation was realized, and the vibration displacements at other positions of a rail were successfully
predicted. Furthermore, an SMEA for guided waves is proposed, through which the single extraction
results of four modes were successfully obtained when the rail was excited along different excitation
directions at a frequency of 200 Hz. In addition, the SMEA was applied to defect location detection,
and the single reflection mode waveform of the defect was extracted. Based on the group velocity
of the mode and its propagation time, the distance between the defect and the excitation point was
measured, and the defect location was predicted as a result. Moreover, the SMEA was applied to
locate the railhead defect. The detection mode, the frequency, and the excitation method Were selected
through the dispersion curves and modal identification results, and a series of signals of the sampling
nodes were obtained using the three-dimensional finite element software ANSYS. The distance
between the defect and the excitation point was calculated using the SMEA result. When compared
with the structure of the simulated model, the errors obtained were all less than 0.5 m, proving the
efficacy of this method in precisely locating rail defects, thus providing an innovated solution for rail
defect location.
Keywords: rail; ultrasonic guided wave; semi-analytical finite element; single mode extraction
algorithm; defect location
1. Introduction
Ultrasound guided waves are widely used in the non-destructive testing of continuous
welded rails because of their wide coverage and rapid propagation over large distances [1,2].
Compared with simple waveguide structures, such as plates and pipelines, rails have much more
complex cross-sectional structures; thus, they require far more complicated ultrasonic wave modes
for defect detection. With the increase in the frequency, the number of modes increases, making the
research more difficult. However, the analytical method is no longer suitable for the analysis of the
propagation characteristics of ultrasonic guided waves in rails. Therefore, numerical analysis methods
have been broadly introduced, including the Boundary Element Method [3], the Finite Difference
Method [4], the Finite Element Method [5], and the Semi-Analytical Finite Element Method [6].
Among them, the semi-analytical finite element method reduces the analysis dimensions and has a
higher computational efficiency, thus is commonly adopted in this field of researches.

Appl. Sci. 2019, 9, 1107
The time difference between the main wave packet of the reflected wave and the excitation
wave packet in the acquisition signal must be calculated; the position of the defect can be inferred by
multiplying the group velocity of the main modes. Its accuracy is related to the selected detection
mode. Therefore, it is necessary to have a good understanding of the propagation characteristics
of guided waves in rails to select the appropriate modal type, frequency, and excitation mode for
defect detection. D. Alleyne and P. Cawley [7] used two-dimensional Fourier transform to separate
the guided wave modes and calculate the corresponding phase velocities.T. Hayashi et al. [8] utilized
the semi-analytical finite element (SAFE) method to calculate the group velocity and phase velocity
dispersion curves of guided waves in rails at frequencies of 0–100 kHz and carried out experimental
verification. He Cunfu et al. [9] used the vibration mode analysis method to analyze the guided wave
modes at frequencies of 0–50 kHz; the dispersion characteristics and wave structures of five typical
modes were analyzed, and the mode types and frequency ranges suitable for inspecting railhead and
rail bottom defects were obtained. P.W. Loveday and C.S. Long [10–14] analyzed the guided wave
mode data in rails at frequencies of 25 kHz and 35 kHz using the SAFE method and measured along
the rails using a laser vibrometer. A large number of vibration data were obtained, the amplitudes of
each mode were calculated, and the vibration waveforms of distant rails were successfully predicted.
In 2017, P.W. Loveday et al. [15] studied the problem of the mode repulsion and crossing behavior
of the approaching wave number versus frequency curves by analyzing the second derivative of the
eigenvalue with respect to the wave number. These methods analyze the modal characteristics of
guided waves propagating in rails, giving a modal analysis. However, it has never been possible to
separate the reflected guided wave modes of defects to use them for defect location directly.
Meanwhile, a large number of research teams have applied ultrasonic guided waves for defect
detection in waveguide structures. Most of the research results focus on simple cross-sectional
structures such as plates and pipes, while studies on complex structures such as rails remain relatively
rare. C.M. Lee et al. [2] adopted the finite element method to analyze the energy distribution
characteristics of guided waves with different frequencies in rails. The simulation results showed that
a high-frequency guided wave of 200 kHz is concentrated on the upper surface of the railhead, and a
low-frequency guided wave of 30 kHz is distributed through the entire railhead. The guided waves at
low frequencies are more sensitive to the transverse cracks of the railhead, while the guided waves at
high frequencies are good at detecting defects in the shelling. These results have been experimentally
verified. Lu Chao et al. [16] selected transverse and vertical vibration modes to detect oblique cracks on
the rail bottom and analyzed the relationship between the angle of the oblique cracks and the scattering
characteristics of the guided waves. G. Zumpano et al. [17] utilized finite element software to simulate
the rail wear and applied different excitation frequencies. Location analysis shows that the location
error is affected by the excitation frequency. However, as a result of the dispersion characteristics of
ultrasonic guided waves and the frequency aliasing phenomenon, it is impossible to accurately locate
the defect from the group velocity. For this reason, there are relatively few studies on the application of
ultrasonic guided waves in rail defects location. Nevertheless, because of the results related with the
prediction of defect sizes and the identification of defect types, it seems practical to study this method
in terms of defect location.
To locate rail cracks accurately by using ultrasonic guided waves, it is generally necessary to
obtain a single mode propagating in the rail. The research ideas can be roughly divided into two
kinds: one is to directly excite a single mode in the rail, and the other is to extract a single mode
from the signal propagating in the rail. In previous studies, the research group proposed a single
mode excitation method, which can excite a relatively pure single mode in rails [18]. At the same
time, the group also carried out research on the optimal mode selection of rail crack detection in
which a selection model of crack detection mode is created. For a specific crack, the guided wave
frequency and mode suitable for crack detection are quickly selected. In this paper, a single mode
extraction algorithm (SMEA) is proposed to ascertain the precise location of defects. The method is
as follows: Firstly, the mode and frequency of defect detection are selected and a three-dimensional
312
Appl. Sci. 2019, 9, 1107
model of the defect present in the railhead is established. The simulation analysis is carried out using
the three-dimensional finite element analysis software ANSYS. The defect models of the railhead
are stimulated with low-frequency (200 Hz) and high-frequency (60 kHz) signals. The content of
each mode of guided wave is quantitatively analyzed, and the position of the railhead defect is
calculated according to the single mode extraction method. The semi-analytical finite element method
and accurate modal identification method are described in Section 2. In Section 3, the SMEA and
verification process are discussed. The defect location method and simulation results are described in
Section 4. Conclusions are given in Section 5.
2. An Accurate Modal Identification Method
2.1. Basic Characteristics of Ultrasonic Guided Waves in Rails

As a result of the complex cross-sectional structure of rails, the number of guided wave modes
propagating in a rail is large. To detect the internal defects of rails based on ultrasonic guided wave
technology, the most important thing is to grasp the fundamental characteristics of ultrasonic guided
waves in rails, such as the frequency, wave number, phase velocity, group velocity, mode shape, and
other information, so as to analyze the propagation characteristics of the guided waves. The dispersion
curves of ultrasonic guided waves in rails can be obtained with the semi-analytical finite element
method. Taking the rail laid on Beijing–Shanghai high-speed railway in China as the research
object [19], the finite element method was used to discretize the cross-section. It was assumed
that the guided waves propagate along the longitudinal direction of the rail in the form of harmonic
vibrations. The wave equation was established based on the finite element method. The eigenvalues
and eigenvectors were obtained by solving the eigenvalue equation. The eigenvalues contain the
information of frequencies and wave numbers, and the eigenvectors contain the information of
mode shapes. Thus, the dispersion curves and mode shapes of guided waves in the CHN60 rail
were obtained.
First, the coordinate system of the CHN60 rail was established, as shown in Figure 1.
Figure 1. CHN60 rail coordinate system.
The wave number of guided waves is ξ and the frequency is ω. The displacement, stress, and strain
of each node in the rail can be expressed as Equation (1).
T
u = ux uy uz
T
σ = σx σy σz σyz σxz σxy (1)
T
ε = εx εy εz γyz γxz γxy
313
Appl. Sci. 2019, 9, 1107
ε x , ε y , and ε z are normal strains, and γyz , γxz , and γxy are shear strains. In the elastic range of
materials, stress and strain conform to Hooke’s Law. It can be expressed as σ = Cε, where C is the
elastic constant matrix of a rail.
The relationship between the strain and displacement at any node in the rail is expressed in a
matrix form as follows:

ε = L x ∂u ∂u ∂u
∂x + Ly ∂y + Lz ∂z (2)
Therefore, in the form

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 0 0 0 0 0 0 0 0
⎢0 0 0⎥⎥ ⎢0 1 0⎥⎥ ⎢0 0 0⎥
⎢ ⎢ ⎢ ⎥
⎢0 0⎥ ⎢0 0⎥ ⎢0 1⎥
⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥
Lx = ⎢ ⎥ Ly = ⎢ ⎥ Lz = ⎢ ⎥ (3)
⎢0 0 0⎥ ⎢0 0 1⎥ ⎢0 1 0⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣0 0 1⎦ ⎣0 0 0⎦ ⎣1 0 0⎦
0 1 0 1 0 0 0 0 0
The SAFE method was used to get the dispersion curves of the ultrasonic guided waves which
propagate in the form of harmonics in the longitudinal direction of the rail. Therefore, only the finite
element discretization of the rail cross-section was needed. The displacement of any discrete node in
the rail can be calculated as shown in Equation (4), where x is the longitudinal coordinate of the rail.
⎡ ⎤ ⎡ ⎤
u x ( x, y, z, t) Ux (y, z)
⎢ ⎥ ⎢ ⎥
u( x, y, z, t) = ⎣ uy ( x, y, z, t) ⎦ = ⎣Uy (y, z) ⎦ e−i(ξx−ωt) (4)
uz ( x, y, z, t) Uz (y, z)
The triangular element was selected to discretize the cross-section of the CHN60 rail,
and 177 nodes and 255 elements were obtained, as shown in Figure 2. Seven nodes are circled in
the figure to illustrate the signal extraction nodes in the subsequent modal identification method.
Figure 2. Discretization of the cross-section of the CHN60 rail.
The rail cross-section was discretized by triangular elements. The displacement of any particle in
an element can be obtained by the displacement of the nodes and the shape function. The strain and
stress vectors of the element can also be expressed by the displacement of the nodes. According to
the Hamilton principle, the dynamic equation of guided waves propagating in a CHN60 rail can be
obtained by calculating the strain energy and potential energy at any point simultaneously [20]:
314
Appl. Sci. 2019, 9, 1107

K1 + iξK2 + ξ 2 K3 − ω 2 M U=0 (5)
M
In this form, U contains the displacements of the nodes in x, y, and z directions. M is the mass
matrix of the nodes. ξ and ω are the wave number and angular frequency, respectively, and K1 , K2 ,
and K3 are the stiffness matrices.
Given the frequency ω, the eigenvalue of Equation (5) is the wave number ξ of the guided waves,
and the eigenvector contains the mode shapes of the rail. Normalized processing can be used to plot
the corresponding mode shapes of the guided waves. The modes with pure imaginary or complex
wave numbers are not considered here because these modes will exponentially decay as the distance
increases and cannot propagate. The dispersion curves of the phase velocity and group velocity of
ultrasonic guided waves in the CHN60 rail are plotted, as shown in Figure 3.
(a) (b)
Figure 3. (a) Phase velocity; and (b) group velocity dispersion curves.
The dispersion curves of the phase velocity and group velocity are shown in the figure. At the
same frequency, the rail has several propagable guided wave modes, and at the higher the frequency,
it has more guided wave modes.
2.2. Excitation Response Analysis of Rails

On the basis of the excitation response analysis method, the vibration signals of a non-defective
rail can be calculated at any point. Firstly, the system function model U of the rail should be
obtained. Then, the frequency domain signal F/( f ) of the excitation function signal v1 (t) can be
obtained.Thereafter, the excitation response result can be calculated V = U × F/( f ). Finally, the inverse
Fourier transform of V can be determined, which is the excitation response of the time domain
vibration signal.
The system function model of the rail [21] is shown in Equation (6), where m is the mode number
(m = 1, 2, . . . , M). Ωm is the amplitude and phase parameters of mode m. Um L and U R are the left
m

eigenvector and the right eigenvector of mode m. p is the amplitude of the excitation signal. xs and x
Rup
are the longitudinal coordinates of the excitation point and the receiving point, and Um is the mode
shape of mode m.
M
∑ Ωm Um e−i[ξ m ( x− xs )]
Rup
U (y, z, f ) = (6)
m =1
In this form, the parameters are expressed as follow:
315
Appl. Sci. 2019, 9, 1107
Um
Lp
Ωm = −
Bm
Bm = Um L R
BUm

K1 − ω 2 M 0
B=
0 − K3
The result of the frequency response analysis of the excitation response V ( f ) can be expressed by
Equation (7).
V (y, z, f ) = F/( f ) · U (y, z, f )

M
= F/( f ) · ∑ Ωm Um e−i[ξ m ( x− xs )]
Rup
(7)
m =1
The inverse Fourier transform result is the time domain vibration signal of the node.
2.3. Modal Identification

In Figure 4, T is the transmitter of the guided waves, R is the receiver of guided waves, X is the
location of a rail defect, and the arrows show the propagation paths of the guided waves. Path 1 is the
direction of backward propagation after exciting the guided wave. Paths 2 and 3 are the directions of
transmission and reflection after the guided wave meets the crack. According to the propagation time
of the reflected wave (Path 3) and the group velocity of the main modes, the guided wave propagating
distance of Path 3 can be calculated to locate the defect.
Figure 4. Schematic diagram of defect location.
To solve the problem of defect location, it is very important to analyze the mode propagation and
reflection of guided waves in defective rails. The amplitude of each mode in a complete rail can be
calculated using the excitation response algorithm, as described in Section 2.2, but the mechanisms
involved in the interaction between modes and defects are still difficult to understand. Therefore,
a method that can accurately analyze the direct mode and reflection mode of a defect is needed.
In fact, the accurate modal identification method used in this paper is the inverse transformation of
the excitation response analysis method.
2.3.1. Theoretical Derivation

According to the excitation response analysis method, the vibration displacement of any point
of the rail is equal to the superposition of all the propagable modes’ vibration displacements at that
point. However, when there are both direct and reflected waves in the rail, the vibration displacement
at any point is the superposition of the direct and reflected waves, that is, the superposition of all
direct modes and all reflected modes at that point. Therefore, when considering the influence of the
reflection wave caused by the defect, the expression of the excitation response of the node is in the form
of Equation (8), in which the upper line “-” represents the relevant parameters of the reflected modes.
316
Appl. Sci. 2019, 9, 1107
V (y, z, f ) = F/( f ) · U (y, z, f )

M M
= F/( f ) · ( ∑ Ωm Um e−i[ξ m ( x− xs )] + ∑ Ω̄m Ūm e−i[ξ̄ m ( x− xs )] )
Rup Rup
(8)
m =1 m =1

If we let Ω = Ωm Ω̄m , then with to the sampling results of a simulation or an experiment,
the frequency domain signal V ( f ) of any N points’ vibration displacement on the rail can be obtained.
By substituting the wave number, the mode shape, the distance between the sampling point and
excitation point, and the frequency domain signals of the excitation signal at the corresponding
frequency of each mode into Equation (8), the parameters Ω including the amplitude and phase of
each mode can be achieved. It should be noted that the number of sampling points should be enough
to ensure that the solution can be obtained relatively accurately, that is, N > 2M.
2.3.2. Simulation Analysis

Due to the excessive modes of high frequency guided waves, the results of modal identification
are complex and difficult to study. Therefore, in the process of research, we first judged the correctness
of the algorithm by studying 200 Hz low-frequency guided waves, and then increased the frequency
to ultrasound band to study the modal identification results of high-frequency guided waves.
Firstly, the rail model with a railhead defect was established, and the excitation response of the
rail was simulated using ANSYS. The rail was then excited by signals at a low center frequency of
200 Hz and a high center frequency of 60 kHz. The rail and defect models are shown in Figure 5.
Figure 5. Rail model with head defect.
(1) Low center frequency of 200 Hz.

The length of the rail model was 200 m, and the defect was located at 120 m and had a length of
70 mm, width of 50 mm, and depth of 20 mm. The excitation signal was a sinusoidal wave modulated
using Hanning window, with a center frequency of 200 Hz and five cycles. The excitation position was
the center point of the side of the railhead at 90 m. A total of 32 sections were sampled from 100 m to
101.55 m. Seven nodes were selected for each section, as shown in Figure 2.
At 200 Hz, there were only four modes in the rail: the horizontal bending mode, the vertical
bending mode, the torsional mode, and the extensional mode. The collected signals were substituted
into the modal identification in Equation (8), which is extended into the form of Equation (9).
317
Appl. Sci. 2019, 9, 1107
⎡ ⎤
U1,1 e−i[ξ 1 (x1 − xs )]
Rup
... U1,4 e−i[ξ 4 (x1 − xs )]
Rup
Ū1,1 e−i[ξ̄ 1 (x1 − xs )]
Rup
... Ū1,4 e−i[ξ̄ 4 (x1 − xs )]
Rup
⎡ V1,1 ( f ) ⎤
⎡ ⎤
⎢ U1,1
Rup −i [ξ ( x − xs )]
e 1 2 ... U1,4 e−i[ξ 4 (x2 − xs )]
Rup
Ū1,1 e−i[ξ̄ 1 (x2 − xs )]
Rup
... Ū1,4 e−i[ξ̄ 4 (x2 − xs )] ⎥ Ω1
Rup
V (f)
⎢ ⎥ Ω2 ⎢ 1,2. ⎥
⎢ .
.
.
.
.
.
.
. ⎥⎢ ⎥ ⎢ . ⎥
⎢ . . . . ⎥⎢ . ⎥ ⎢ . ⎥
⎢U Rup e−i[ξ1 (x32 −xs )] U1,4 e−i[ξ 4 (x32 − xs )]
Rup
Ū1,1 e−i[ξ̄ 1 (x32 − xs )]
Rup
Ū1,4 e−i[ξ̄ 4 (x32 − xs )] ⎥
Rup ⎢.⎥ ⎢V1,32 ( f )⎥
⎢ 1,1 ... ... ⎥⎢ . ⎥ ⎢ ⎥
⎢ ⎥ ⎢ Ω4 ⎥ ⎢ ⎥
F/( f ) ⎢ .
.
.
.
.
.
.
. ⎥ ⎢Ω̄1 ⎥ = ⎢ ... ⎥ (9)
⎢ Rup −i[.ξ (x −x )] . . . ⎥⎢ ⎥ ⎢ ⎥
⎢ U7,1 e 1 1 s ...
Rup −i [ξ ( x − xs )]
U7,4 e 4 1
Rup −i [ξ̄ ( x − xs )]
Ū7,1 e 1 1 ... Ū7,4 e−i[ξ̄ 4 (x1 − xs )] ⎥ ⎢ Ω̄2 ⎥
Rup
⎢ V7,1 ( f ) ⎥
⎢ Rup −i[ξ1 (x2 −xs )] Rup −i [ξ̄ ( x − xs )] ⎥ ⎢ . ⎥ ⎢ V7,2 ( f ) ⎥
⎢ U7,1 e ...
Rup −i [ξ ( x − xs )]
U7,4 e 4 2
Rup −i [ξ̄ ( x − xs )]
Ū7,1 e 1 2 ... Ū7,4 e 4 2 ⎥⎣ . ⎦ ⎢ ⎥
⎢ ⎥ . ⎣ .. ⎦
⎣ .
.
.
.
.
.
.
.
.
.
.
.
⎦ Ω̄4 .
Rup −i [ξ ( x − xs )]
U7,1 e 1 32 ...
Rup −i [ξ ( x − xs )]
U7,4 e 4 32
Rup −i [ξ̄ ( x − xs )]
Ū7,1 e 1 32 ... Ū7,4 e−i[ξ̄ 4 (x32 − xs )]
Rup V7,32 ( f )
The magnitudes of each mode and the corresponding reflection mode (i.e., the modulus value of
Ω) and the amplitude reflection coefficients after calculation are shown in Table 1.
Table 1. Vertical excitation mode identification results of defective rail (at a frequency of 200 Hz).
Mode Direct Wave Reflected Wave Amplitude Reflection Coefficient

Horizontal bending mode 3.20 × 10−3 3.04 × 10−4 9.5 × 10−2
Vertical bending mode 1.89 × 10−2 3.52 × 10−3 0.19
Torsional mode 1.17 × 10−2 4.58 × 10−4 3.9 × 10−2
Extensional mode 4.94 × 10−4 5.95 × 10−4 1.2
It shows that the vertical bending mode had the highest amplitude among the direct modes,
which was the main mode to be stimulated in this way. The amplitude of the extensional mode
was far smaller than the amplitudes of the other modes and could be seen as zero, thus it was not
considered. Meanwhile, the amplitude reflection coefficient was obtained by dividing the amplitude of
the reflected mode from that of the direct mode, and the vertical bending mode was the most sensitive
mode to the transverse crack of the railhead. The results of the modal identification algorithm were
used to estimate the vibration of the railhead at a distance of 20 m away from the excitation point
and were compared with the simulation sampling results at the same point, as shown in Figure 6a.
The two methods’ results almost coincide, which proves the validity of the algorithm in the modal
identification of guided waves at a low frequency.
(2) High center frequency of 60 kHz
The length of the rail model was 25 m, and the defect was located at 20 m, with a length of 70 mm,
a width of 50 mm, and a depth of 20 mm. A 10-cycle sinusoidal wave with a center frequency of
60 kHz was applied to excite the railhead at 12 m vertically, and 32 sections were sampled between
12.3 m and 13.23 m. The nodes were selected as above.
At 60 kHz, there were 35 modes in the rail, thus it was impossible to simply distinguish the
types of modes. For this reason, they are numbered in the order of phase velocity from small to large.
Using the signal with a center frequency of 60 kHz to excite the railhead, the magnitude and reflection
coefficients of each mode and the corresponding reflection modes were obtained (as shown in Table 2),
after the application of the accurate modal identification method.
In Table 2, it can be seen that, among the direct modes, mode No. 7, No. 8, and No. 9 had
relatively high amplitudes, as they were the main modes stimulated in this way. At the same time,
in the reflection mode, mode No. 7, with the largest amplitude and the highest reflection coefficient,
was the most sensitive mode to the transverse crack of the railhead. Similarly, the vibration of the
railhead 1.5 m away from the excitation point was estimated using the calculation results in Table 2,
and compared with the simulation sampling results at that point, as shown in Figure 6b. The two
main wave packets almost coincide, which shows that the algorithm is also applicable to the modal
identification of guided waves at a high frequency.
318
Appl. Sci. 2019, 9, 1107
(a) (b)
Figure 6. Comparison between the simulation results and prediction results: (a) 200 Hz; and (b) 60 kHz.
Table 2. Results of accurate modal identification (at a frequency of 60 kHz).
Mode Number Direct Wave Reflected Wave Amplitude Reflection Coefficient

1 0.39 0.003 0.008
2 0.18 0.009 0.050
3 0.03 0.001 0.033
4 0.29 0.007 0.024
5 0.33 0.003 0.009
6 0.21 0.008 0.038
7 1.25 0.078 0.062
8 1.46 0.004 0.003
9 1.36 0.004 0.003
10 0.52 0.002 0.004
11 0.08 0.001 0.013
12 0.23 0.002 0.009
13 0.40 0.003 0.043
14 0.67 0.001 0.001
15 0.07 0.003 0.043
16 0.06 0.001 0.017
17 0.96 0.005 0.005
18 0.04 0 0
19 0.37 0.006 0.016
20 0.02 0.001 0.050
21 0.29 0.005 0.017
22 0.15 0.002 0.013
23 0.06 0.001 0.017
24 0.16 0.003 0.019
25 0.23 0.008 0.035
26 0.28 0.001 0.004
27 0.36 0.002 0.006
28 0.41 0.001 0.002
29 0.42 0.002 0.005
30 0.03 0.001 0.033
31 0.32 0 0
32 0.42 0 0
33 0.92 0 0
34 0.02 0 0
35 0.11 0 0
319
Appl. Sci. 2019, 9, 1107
3. Single Modal Extraction Algorithm

When ultrasonic guided waves propagate in rails, dispersion occurs. That is to say, the number of
the guided wave modes propagating in the rails will increase with the increase in excitation frequency;
furthermore, the propagating velocity of each mode is different. With the increase in propagating
distance, the wave packets of each mode will gradually stagger and overlap. Therefore, they cannot
be distinguished, and so the group velocity cannot be used to locate defects. To solve this problem,
an SMEA based on the accurate modal identification method is proposed.
As mentioned above, the vibration displacement at any point on the rail is the superposition of
the vibration displacement of all modes at this point. Conversely, the vibration displacement of each
mode at any point can be split by the total displacement of the point.
The amplitude and phase Ωn of mode n can be obtained with the accurate modal identification
algorithm, and the vibration displacement generated by mode n at the longitudinal coordinate x can
be expressed by Equation (10).
Rup −i [ξ n ( x − xs )]
Vn (y, z, f ) = F̄f · Ωn Un e (10)
According to Equation (10), the vibration displacement of mode n at this point in a frequency
domain is calculated, and the time domain waveform of the mode is obtained by inverse Fourier
transform. To verify the correctness of the method, the following simulations were carried out.
A non-defective three-dimensional rail model with a length of 200 m was established.
The vibrations of the railhead along the rail in a longitudinal, transverse, and vertical excitation
direction were simulated using ANSYS. To show the results clearly, the excitation signal was selected
as a low-frequency signal with a center frequency of 200 Hz and five cycles, which has only four
modes. The modal amplitude of each mode at the 200 Hz frequency point after mode analysis is shown
in Figure 7.
Figure 7. Modal identification results under three excitation conditions (200 Hz).
According to Figure 7, when the rail was excited longitudinally, the proportion of the extensional
mode was higher. When the rail was excited horizontally, the proportions of the horizontal bending
mode and torsion mode were higher. When the rail was excited vertically, the proportions of the
vertical bending mode and torsion mode were higher.
Using the SMEA and choosing the frequency range 100–300 Hz, the total waveform at 30 m and
four single-mode waveforms were obtained, respectively, as shown in Figure 8a–c.
It can be seen clearly in the figure that only the extensional mode existed when the railhead was
excited longitudinally. The horizontal bending and torsion modes existed when the railhead was
excited horizontally, and the vertical bending and torsion modes existed when the railhead was excited
vertically. The results are consistent with the accurate identification results.
320
Appl. Sci. 2019, 9, 1107
(a) (b) (c)

Figure 8. SMEA results: (a) longitudinal excitation; (b) horizontal excitation; and (c) vertical excitation.
4. Defect Location
The flow chart of the defect location method is shown in Figure 9.
Figure 9. Defect location algorithm flow chart.
As can be seen above, the primary task of defect location is to select the mode, frequency,
and excitation conditions with which to perform the defect detection. At the same time,
a three-dimensional model of the rail with defects must be established. The excitation response
of the rail is then simulated by ANSYS, and the vibration displacements of a series of points on the
rail are obtained, so as to simulate the signals received by an experiment. According to the results
of the modal identification and the SMEA, the reflected signals of the selected modes are plotted.
Thereafter, the group velocity and the propagation time of the reflected mode are obtained, so that
the defect location can be achieved. The following section takes a transverse crack in a railhead as an
example to explain the process of defect location.
4.1. Selection of Mode, Frequency, and Excitation Conditions for Defect Detection
The following three principles are used to select the mode, frequency, and excitation conditions
for railhead defect detection:
• The mode that only vibrates in the railhead with almost no movement of rail waist and rail bottom
and which has a large group velocity is selected.
321
Appl. Sci. 2019, 9, 1107
• The frequency band with better non-dispersive characteristics is selected.

• The mode with the largest amplitude is selected as the excitation condition.
The detection frequency and mode were selected in the frequency range 20–70 kHz. According to
Equation (5), the mode shapes of the rail in this frequency range were calculated and drawn.
The frequency of 60 kHz was taken as an example, as shown in Figure 10. Among them, the modes
in which only the railhead vibrated with almost no movement in the rail waist and rail bottom were
modes No. 7 and No. 14.
Figure 10. Rail mode shapes (60 kHz).
According to Figure 3b, the group velocity dispersion curves of modes No. 7 and No. 14 were
extracted, as shown in Figure 11.
Figure 11. Group velocity dispersion curves of modes No. 7 and No. 14.
322
Appl. Sci. 2019, 9, 1107
In Figure 11, we can see that mode No. 7 had almost the same group velocity and excellent
non-dispersive characteristics in the proximity of 60 kHz. The group velocity of mode No. 14 was
greatly influenced by the frequency and the non-dispersion characteristic was poor. Therefore, mode
No. 7 at 60 kHz frequency was selected to detect the defect in the railhead.
A complete three-dimensional rail model with a length of 15 m was established. A simulation
whereby the railhead was excited at 7 m along the longitudinal, transverse, and vertical directions by
ANSYS was performed. The excitation signal was a sinusoidal wave modulated by Hanning window
at the center frequency of 60 kHz with 10 cycles. The co-directional displacement signals between
8 m and 9 m were collected, and the amplitude of each mode under three excitation conditions were
calculated using the accurate modal identification method, as shown in Figure 12.
As shown in Figure 12, the highest amplitude for the selected mode is that of vertical excitation.
Therefore, vertical excitation was selected.
In summary, mode No. 7 was selected as the detection mode, 60 kHz was selected as the frequency,
and vertical excitation was selected as the excitation condition.
Figure 12. Modal identification results under three excitation conditions (60 kHz).
4.2. Simulation Analysis of Defect Location

According to the results in Section 4.1, the rail with a defect in the railhead was stimulated.
As shown in Figure 13, the rail length was 25 m; T represents the excitation point, which was located at
12 m along the rail. X represents the defect position, and the distance between the excitation point and
X was l = 8 m. The simulation condition used was the same as the high-frequency excitation presented
in Section 2.3.2. Table 2 shows the results of the modal identification.
Figure 13. Schematic diagram of rail defect location.
The spacings d between the sampling points and the excitation point were 1.5 m, 4.5, m and 6 m,
and the collected signal waveforms are shown in Figure 14a–c.
323
Appl. Sci. 2019, 9, 1107
(a) (b) (c)

Figure 14. Acquisition waveforms for all modes with distances of: (a) 1.5 m; (b) 4.5 m; and (c) 6 m
between excitation points and sampling points.
As a result of the complex propagation modes of guided waves in rails, only the first direct wave
packet could be distinguished from the waveform. It was impossible to pick out which wave packet in
the red circle was the reflected wave packet, and it was also impossible to know for certain which mode
or modes each wave packet corresponded to; therefore, it was difficult to give the specific propagation
time and the accurate mode group velocity. For this reason, the defect could not be precisely located.
According to the SMEA, the reflected mode waveform of mode No. 7 at 4.5 m from excitation
point was calculated in the frequency range 42–78 kHz. As shown in Figure 15b, the peak time t1 of the
wave packet was 3.966 × 10−3 s. With the peak time t0 of the excitation wave packet (8.333 × 10−5 s)
and the group velocity v0 of mode No. 7 (3148.2 m/s), it was possible to calculate the distance between
the defect and the excitation point using Equation (11).
( t1 − t0 ) ∗ v0 + d
l= (11)
2
The actual interval was 8 m and the calculation result was 8.36 m, giving an error of only 0.36 m.
The reflection mode waveforms of mode No. 7 at 1.5 m and 6 m away from the excitation point
were extracted, as shown in Figure 15a,c. The distances between the calculated defect and the excitation
point were 8.33 m and 8.39 m. Therefore, the errors were 0.33 m and 0.39 m, respectively, both less
than 0.5 m, and thus meeting the positioning accuracy requirements.
(a) (b) (c)

Figure 15. The reflection waveforms of mode No. 7 at a distance from the excitation point of: (a) 1.5 m;
(b) 4.5 m; and (c) 6 m.
5. Conclusions
A rail defect location method based on an SMEA of ultrasonic guided waves was proposed.
The dispersion curves of a CHN60 rail were calculated using the semi-analytical finite element method,
the mode shapes of the guided waves were calculated, and the actual received signals were predicted
using the simulation data provided by ANSYS. Furthermore, accurate modal identification was
achieved using the inverse transformation of the excitation response method, and the amplitude
of each mode of the guided waves propagating in the rail was obtained. The rail displacement
324
Appl. Sci. 2019, 9, 1107
curves excited by the signals at a low frequency of 200 Hz and high frequency of 60 kHz were
correctly predicted.
According to the modal identification results, an SMEA was proposed. The single mode extraction
results at 200 Hz were calculated and the results were consistent with those of the modal identification.
Thereafter, taking a railhead defect as an example, according to certain selection principles, mode No. 7
with a signal frequency of 60 kHz and a vertical excitation condition were selected. The reflective
modes of mode No. 7 at 1.5, 4.5, and 6 m from the excitation point of the railhead defect were extracted
using the SMEA, and the distance between the rail defect and excitation point was obtained according
to the time difference. The errors in location were 0.33, 0.36, and 0.39 m, respectively, which all fall
within the 0.5 m set as the detection requirements. Hence, it can be said that the precise determination
of the rail defect was accomplished.
Author Contributions: Conceptualization, B.X. and X.X.; methodology, X.X. and L.Z.; software, B.X. and H.S.;
validation, Z.Y., L.Z. and H.S.; investigation, B.X.; resources, Z.Y.; writing—original draft preparation, B.X.; and
writing—review and editing, B.X., X.X. and L.Z.
Funding: This research was funded by the National Key Research and Development Program of China
(2016YFB1200401)
Abbreviations
The following abbreviations are used in this manuscript:
SAFE Semi-analytical finite element

SMEA Single mode extraction algorithm
References
1. Loveday, P.W. Guided wave inspection and monitoring of railway track. J. Nondestruct. Eval. 2012,
31, 303–309. [CrossRef]
2. Lee, C.; Rose, J.L.; Cho, Y. A guided wave approach to defect detection under shelling in rail. NDT E Int.
2009, 42, 174–180. [CrossRef]
3. Rose, J.L.; Zhu, W.; Cho, Y. Boundary element modeling for guided wave reflection and transmission factor
analyses in defect classification. In Proceedings of the 1998 IEEE Ultrasonics Symposium, Sendai, Japan,
5–8 October 1998; Volume 1, pp. 885–888.
4. Harker, A.H. Numerical modelling of the scattering of elastic waves in plates. J. Nondestruct. Eval. 1984,
4, 89–106. [CrossRef]
5. Moser, F.; Jacobs, L.J.; Qu, J. Modeling elastic wave propagation in waveguides with the finite element
method. NDT E Int. 1999, 32, 225–234. [CrossRef]
6. Gavrić, L. Computation of propagative waves in free rail using a finite element technique. J. Sound Vib. 1995,
185, 531–543. [CrossRef]
7. Alleyne, D.; Cawley, P. A two-dimensional Fourier transform method for the measurement of propagating
multimode signals. J. Acoust. Soc. Am. 1991, 89, 1159–1168. [CrossRef]
8. Hayashi, T.; Song, W.; Rose, J.L. Guided wave dispersion curves for a bar with an arbitrary cross-section,
a rod and rail example. Ultrasonics 2003, 41, 175–183. [CrossRef]
9. He, C. Propagation characteristics of ultrasonic guided wave in rails based on vibration modal analysis.
J. Vib. Shock 2014, 33, 9–13. [CrossRef]
10. Loveday, P.W.; Long, C.S. Laser vibrometer measurement of guided wave modes in rail track. Ultrasonics
2015, 57, 209–217. [CrossRef] [PubMed]
11. Loveday, P.W. Measurement of modal amplitudes of guided waves in rails. Proc. SPIE 2008, 6935, 1J-1–1J-8.
[CrossRef]
12. Loveday, P.W. Modeling and Measurement of Piezoelectric Ultrasonic Transducers for Transmitting Guided
Waves in Rails. In Proceedings of the 2008 IEEE Ultrasonics Symposium, Beijing, China, 2–5 November 2008;
pp. 410–413.
325
Appl. Sci. 2019, 9, 1107
13. Loveday, P.W.; Long, C.S. Modal amplitude extraction of guided waves in rails using scanning laser
vibrometer measurements. Am. Inst. Phys. Conf. Ser. 2012, 1430, 182–189. [CrossRef]
14. Loveday, P.W.; Long, C.S. Field measurement of guided wave modes in rail track. Am. Inst. Phys. Conf. Ser.
2013, 1511, 230–237. [CrossRef]
15. Loveday, P.W.; Long, C.S.; Ramatlo, D.A. Mode repulsion of ultrasonic guided waves in rails. Ultrasonics
2017, 84, 341–349. [CrossRef] [PubMed]
16. Chao, L.U.; Sheng, H.J.; Song, K.; Lin, J.M.; He, F. Ultrasonic Guided Wave Scattering Characteristics of Rail
Base Oblique Cracks. Nondestruct. Test. 2016, 38, 18–25. [CrossRef]
17. Zumpano, G.; Meo, M. A new damage detection technique based on wave propagation for rails. Int. J.
Solids Struct. 2006, 43, 1023–1046. [CrossRef]
18. Xu, X.; Zhuang, L.; Xing, B.; Yu, Z.; Zhu, L. An Ultrasonic Guided Wave Mode Excitation Method in Rails.
IEEE Access 2018, 6, 60414–60428. [CrossRef]
19. Jian, W.; Ping, W.; Ma, D. Matching performance of CHN60N rail with various wheel profiles for Chinese
high-speed railway. IOP Conf. Ser. Mater. Sci. Eng. 2018, 383, 012041. [CrossRef]
20. Zhang, Z.-T.; Yan, L.-S.; Wang, P.; Guo, L.-K.; Pan, W.; Zhang, Z.-Y. Key Techniques for Rail Strain
Measurements Based on Fiber Bragg Grating Sensor. J. China Railw. Soc. 2012, 34, 65–69. [CrossRef]
21. Bartoli, I.; Coccia, S.; Phillips, R.; Srivastava, A.; Scalea, F.L.D.; Salamone, S.; Fateh, M.; Carr, G. Stress
dependence of guided waves in rails. Proc. SPIE 2010, 7650, 765021. [CrossRef]
326
applied
sciences
Article
Numerical and Experimental Investigation of Guided
Wave Propagation in a Multi-Wire Cable
Pengfei Zhang 1 , Zhifeng Tang 2, *, Fuzai Lv 1 and Keji Yang 1
Zhejiang University, 38 Zheda Road, Hangzhou 310027, China; zhangpengfei@zju.edu.cn (P.Z.);
lfzlfz@zju.edu.cn (F.L.); yangkj@zju.edu.cn (K.Y.)
2 Institute of Advanced Digital Technologies and Instrumentation, College of Biomedical Engineering &
Abstract: Ultrasonic guided waves (UGWs) have attracted attention in the nondestructive testing
and structural health monitoring (SHM) of multi-wire cables. They offer such advantages as a single
measurement, wide coverage of the acoustic field, and long-range propagation ability. However,
the mechanical coupling of multi-wire structures complicates the propagation behaviors of guided
waves and signal interpretation. In this paper, UGW propagation in these waveguides is investigated
theoretically, numerically, and experimentally from the perspective of dispersion and wave structure,
contact acoustic nonlinearity (CAN), and wave energy transfer. Although the performance of all
possible propagating wave modes in a multi-wire cable at different frequencies could be obtained
by dispersion analysis, it is ineffective to analyze the frequency behaviors of the wave signals of
a certain mode, which could be analyzed using the CAN effect. The CAN phenomenon of two
mechanically coupled wires in contact was observed, which was demonstrated by numerical guided
wave simulation and experiments. Additionally, the measured guided wave energy of wires located
in different layers of an aluminum conductor steel-reinforced cable accords with the theoretical
prediction. The model of wave energy distribution in different layers of a cable also could be used to
optimize the excitation power of transducers and determine the effective monitoring range of SHM.
Keywords: guided wave; multi-wire cable; wave structure; contact acoustic nonlinearity;
energy transfer
1. Introduction
Multi-wire cables play an important role as the main structural components in such mechanical
and civil engineering applications as load carriers in elevators, lifting machinery, and cable-stayed and
suspension bridges; post-tensioners in civil structures; and as overhead transmission lines (OVTLs) in
power grids. Nondestructive inspection and structural health monitoring (SHM) of these cables
are crucial to ensure the proper structural performance of cable-stayed and suspension bridges
and high-voltage transmission towers. Bridge cables consist of multi-layer steel wires arranged
in parallel within an equilateral hexagon and a polyethylene sheath wrapped around the wire bundle.
Aluminum conductor steel-reinforced (ACSR) cables are the main components of the OVTLs between
the transmission towers.
Many nondestructive testing methods are available, such as visual inspection [1,2], radiography [3],
computed tomography [4], acoustic emission monitoring [5], ultrasound [1,6], and magnetic flux
leakage [7]. However, many of these methods have limitations when applied to multi-wire cables.
For instance, conventional ultrasonic testing methods are impracticable for detecting large and long
structures, such as multi-wire cables, owing to the deficiency of local detection. Ultrasonic guided

Appl. Sci. 2019, 9, 1028
wave (UGW)-based nondestructive testing (NDT) techniques are increasingly being used to inspect
various multi-wire cables given their high sensitivity, long-range inspection, and full cross-sectional
coverage properties. Many studies [8–18] related to guided wave dispersion properties, propagation
characterization, signal processing methods, and wire breakage detection in multi-wire cables have
been reported.
When UGWs encounter the end or defect of a waveguide in propagation, reflected waves
propagate along the original wave path and are received by the transducer, and the transmitted
waves continue to propagate forward. Defect determination methods are used to measure the size
and location of the defect, and evaluate the structural health state. The cross-section of the waveguide
confines UGWs to propagate in only the length direction. However, the dispersive and multimodal
nature of UGW propagation has to be analyzed. UGWs with a narrow frequency bandwidth often
consist of several superposed modes with different wave structures [19] and propagation velocities.
Therefore, the appropriate signal processing method is needed to analyze the acquired signals of
received waves [20]. Multi-wire cables are composed of a number of parallel (e.g., bridge cable) or
twisted (e.g., steel wire strands, ACSR cable) wires. The factors of helical and twisted geometry,
tension state, and contact and friction stresses between adjacent wires complicate the numerical
models of multi-wire structures [21]. Cables are commonly composed of a number of twisted wires
with different diameters and materials, resulting in different wave velocities even at the same center
excitation frequency. The contact condition between adjacent wires contributes to the complexity of
the dispersion and propagation characteristics of the multi-wire structures. The factors of a large
number of wires, a small diameter of an individual wire, and long length present challenges to the
numerical modeling of guided waves propagation in multi-wire cables, such as slow convergence
and time-consuming computation. The cable sheathing and irregular wire bundle surface geometry
also make it difficult to design a transducer with high coupling efficiency, which is a critical part in
industrial applications in the field.
The overall dispersion characteristics of a multi-wire system with consideration of the contact
stress between adjacent wires have been studied using a double-rod model, in which the guided wave
mode conversion phenomenon was verified [22]. The dispersion curves could display the propagation
velocities of all possible guided wave modes at different frequencies in a structure. However, as for
a UGW with a given mode and center excitation frequency propagating in a multi-wire structure,
dispersion curves are incapable of predicting the change in frequency components of the wave signals
due to the mechanical contact of adjacent wires.
Studies of nonlinear guided waves used to detect microcracks in structures and to characterize
materials have been reported [23,24]. Higher-order harmonic waves were generated in the structure
with microcracks when conducting the NDT inspection using an ultrasonic wave. Similarly, the contact
acoustic nonlinearity (CAN) effect is also in existence in a contact multi-wire system. A longitudinal
wave is excited in one rod of the two-contact-rod system, and harmonic waves with a frequency twice
the excitation frequency are generated in both of the rods due to the CAN effect.
Piezoelectric and magnetostrictive transducers have been widely used to excite and receive
guided waves in a multi-wire cable. However, there is a nonmagnetostrictive effect in aluminum
wires of an ACSR cable, and the magnetostriction of the innermost steel wires is too weak to generate
guided waves due to the influence of the lift distance on the bias magnetic field caused by the layers
of aluminum wires. Therefore, piezoelectric transducers, as practicable transducing devices, are
commonly employed to carry out an inspection of an ACSR cable using UGWs. Several piezoelectric
transducers, small in size, are uniformly installed on the outermost aluminum wires of the cable at
the circumference [25]. The uniform coverage of the guided waves in the cable section relies on the
contact between adjacent wires to transfer waves into inner layers from the outermost layer, where
the transducers are installed. Models of the distribution and propagation of guided wave energy
in different layers could provide guidance for determining the excitation power of transducers and
estimating the monitoring range of the SHM system.
328
Appl. Sci. 2019, 9, 1028
In this paper, UGW propagation in a single cylindrical waveguide and double mechanically
coupled cylindrical waveguides is introduced, including the dispersion properties and CAN effect.
There are many different wave modes simultaneously propagating in a structure with different
velocities at different frequencies or even the same frequency, which are partially caused by
the dispersion phenomenon. However, the appearance of the high-order harmonic wave in the
signal of a single mode wave with a single excitation frequency is caused by the CAN effect.
An efficient energy-based model of coupled cylindrical waveguides is used to predict the wave
energy distribution in wires located in different layers of an ACSR cable, by which the uniformity
of the cross-sectional distribution of the guided wave could be obtained. Numerical transient finite
element (FE) simulation and a two-dimensional (2D) fast Fourier transform were used to verify the
results of the theoretical analysis. Experimental investigations were completed to generate and receive
UGWs in a two-contact-wire cable and an ACSR cable with five-layer wires. The proposed numerical
simulation models were fit to the experimental data.
2. UGW Propagation in Multi-Wire Waveguides

The mechanical properties of multi-wire cables are more complicated than those of individual
solid cylindrical rods. The most prominent features are the complex geometry of the cross-section and
the contact and friction stresses between adjacent wires, which are largely caused by the tension loads
during service. The preliminary understanding of UGW propagation in a multi-wire cable was used to
study the wave propagation behavior of an individual wire using the analytical Pochhammer–Chree
equations [26]. An infinite number of guided wave modes, which are classified by their particle
displacement, might propagate through the structure. These modes in a cylindrical waveguide are
composed of the longitudinal waves L(m,n) with radial and axial vibration displacement, torsional
waves T(m,n) with circumferential vibration displacement, and flexural waves F(m,n) with radial, axial,
and circumferential vibration displacement, where m ∈ {0,1,2 . . . } denotes the circumferential order
and n ∈ {1,2,3 . . . } stands for the nth root of the characteristic equation [27].
The performance of all possible propagating wave modes in a multi-wire cable at different
frequencies could be predicted by UGW dispersion curves. However, when a guided wave was excited
in one of the contact wires with a single-frequency excitation signal, it is ineffective to analyze the
frequency behaviors of the wave signals for a certain mode, which could be analyzed using the CAN
effect. From the perspective of UGW energy transfer, the distribution of wave energy in each layer of a
cable composed of multiple layers of wires with different materials could be obtained.
2.1. Dispersion Analysis of a Single Rod

A single wire can be considered to be the same as an individual cylindrical rod, whose dispersion
properties can be analytically obtained using the Pochhammer–Chree equations. UGWs tend to travel
with different velocities in a structure depending on the center excitation frequency. There are two
ways to define the propagation velocities of UGWs: group and phase velocity. Group velocity refers
to the propagation velocity of the entire wave packet. Phase velocity is the propagation velocity at
which the phase of any one frequency component of the wave travels. For a cylindrical rod, group and
phase velocity dispersion curves of a 7-mm-diameter steel wire (commonly used in a bridge cable),
and a steel and aluminum wire with a diameter of 2.5 mm and 3.2 mm, respectively (commonly used
in ACSR cable), were acquired using an open source software package named PCDISP developed in
the Matlab (MathWorks Inc., Natick, MA, USA) environment. PCDISP is described in more detail in
references [28,29]. The material and geometric properties of wires used in the dispersion calculation
models are summarized in Table 1. The results are shown in Figure 1.
329
Appl. Sci. 2019, 9, 1028
Table 1. The material and geometric properties of wires used in the dispersion models.
Material Diameter Elastic Modulus Density Poisson Ratio

Steel 7 mm 209 GPa 7800 kg/m3 0.3
Steel 2.5 mm 209 GPa 7800 kg/m3 0.3
Aluminum 3.2 mm 72 GPa 2700 kg/m3 0.35

) /
)
/ /

7 /
*URXS YHORFLW\PV

3KDVH YHORFLW\PV
/
7 / )
)
7
)
) )

)
7

)UHTXHQF\N+] )UHTXHQF\N+]
(a)

/

*URXS YHORFLW\ PV

3KDVH YHORFLW\PV
/

7

7
)

)

6WHHOZLUHD=2.5mm 6WHHOZLUHD=2.5mm
$OXPLQXPZLU HD=3.2mm $OXPLQXPZLU HD=3.2mm

)UHTXHQF\N+] )UHTXHQFHN+]
(b)
Figure 1. Phase and group velocity dispersion curves for (a) a single 7-mm-diameter steel wire and (b)
a single steel and aluminum wire with a diameter of 2.5 mm and 3.2 mm, respectively.
Figure 1 shows UGW modes for three different types of wires with frequencies up to 1000 kHz.
The phase and group velocities of the fundamental order flexural mode F(1,1) are slow and display
severe dispersive properties in the low-frequency range (below 200 kHz). L(0,1) and T(0,1) seem to be
the applicable modes for wire breakage inspection of steel and aluminum wires, though the velocity
of the L(0,1) mode is faster than that of the T(0,1) mode at low frequencies, yet is slightly dispersive.
Therefore, the L(0,1) mode seems ideal for wire breakage detection due to its velocity and relatively
low dispersive behavior. Figure 2a–c show the vibration displacement distribution in a cross-section of
a single 7-mm-diameter steel wire carrying these three types of fundamental modes: L(0,1), T(0,1), and
F(1,1), at a frequency of 60 kHz. The arrows indicate the particle vibration displacements in the radial
direction (in-plane). The color indicates the magnitude of the particle displacements in the z-axial
direction (out-of-plane).
330
Appl. Sci. 2019, 9, 1028
(a) (b) (c)

N+]
N+]
N+]
N+]
5DGLXV PP
5DGLXV PP
N+]
8]
8]
8U
N+] N+]
N+] N+]
N+] N+]
N+] 8U N+]
N+] N+]

'LVSODFHPHQW1RUPDOL]HG 'LVSODFHPHQW1RUPDOL]HG
(d) (e)
Figure 2. The vibration displacement distribution of (a) longitudinal, (b) torsional, and (c) flexural
waves in a 7-mm-diameter steel wire. Wave structure of the (d) L(0,1) and (e) F(1,1) modes at
different frequencies.
Both longitudinal and flexural guided waves create radial and axial vibration displacement.
The coupling extent among the adjacent contact wires in the cable is determined by the radial
vibration displacement at the wire surface (i.e., maximum radius). The larger the radial vibration
displacement amplitude, the better the coupling. Therefore, the frequency of UGW wave modes with
large radial vibration displacement at the wire surface become particularly critical during energy
transfer. In Figure 2d,e, radial and axial vibration displacement distribution curves (wave structure) of
the different radii of a 7-mm-diameter steel wire were calculated using Pochhammer–Chree equations
for the L(0,1) and F(1,1) wave modes in the 20–180 kHz frequency range. As the wire radius increases,
the radial displacements decrease but the axial displacements increase. This trend was significant with
the increase in wave frequency below 200 kHz.
2.2. Contact Acoustic Nonlinearity of Coupled Double Rods

Higher harmonic ultrasonic waves are generated in a structure with imperfect interfaces, in which
excited waves with a large amplitude are incident. This phenomenon is called the contact acoustic
nonlinearity (CAN) effect [24]. This phenomenon is caused by the repeated reflection of incident
waves between two surfaces. The same behavior occurs when a longitudinal UGW propagates to
an imperfect interface formed by two surfaces of the contact structures. The compressional part of
the wave can penetrate through the interface and propagate into another structure, but the tensile
part cannot. Therefore, it is equivalent to a half-wave rectification of the transmitted wave, which is a
nonlinear effect. This nonlinearity is embodied in the form of higher harmonics. In nondestructive
inspection using a guided wave, when the characteristic length of the rough contact surfaces or a
micro-crack is much smaller than the incident wavelength, a nonlinear phenomenon occurs.
331
Appl. Sci. 2019, 9, 1028
In a pioneering work [30], the second harmonics generated by the nonlinear nature of the contact
stiffness with the frequency of 2f was theoretically observed in both transmitted and reflected waves,
where f is the frequency of incident waves. A pair of coupled waveguides were created with an active
wire and a passive wire closely in contact with each other. An ultrasonic guided wave was excited on
the left end of the active wire. The radial displacements of the active wire were partially coupled to
the passive wire through contact stress in the contact interface. Additional waves were generated in
the active and passive wires as a result of the local load provided by the contact stress, as shown in
Figure 3.
FRPSUHVVLRQDOSDUW X]
Guided
WHQVLOHSDUW XU
wave
f inc = f
f inc = f $FWLYHZLUH
fref = 2f
ftra = 2f
3DVVLYHZLUH
Figure 3. A schematic drawing of contact acoustic nonlinearity (CAN) at the rough contact interface of
a pair of coupled cylindrical waveguides.
2.3. Energy Model of Multi-Wire Waveguides

There are many limitations of modeling accurate UGW propagation of multi-wire structures
taking into account the effect of mechanical contact, such as varying contact stress among the adjacent
wires, wave mode conversion, and the CAN effect. In order to simplify the propagation model of an
UGW in a multi-wire cable, a model that describes the transfer of the wave between an active and a
passive wire from an energy perspective was used to analyze the propagation behavior of an UGW.
A multi-wire system was composed of mechanically coupled wires. They were surrounded by a spring
array along the wire axis [17,18]. Thus, the wave energy of the ith wire of a cable with n wires can be
mathematically described by a second-order nonlinear differential equation:

d2 Ei (z) dE (z) n
= −dm i −c Ei (z) ∑ bij ( Ei (z) − Ej (z)) (1)
dz2 dz j =1
where Ei (z) is the average energy transfer from all n wires to the ith wire, dm is the damping coefficient
caused by the material, c is the mechanical coupling coefficient, and a binary constant bij indicates
whether the wires i and j are in contact with each other, expressed as:

1 in contact
bij = . (2)
0 otherwise
For a two-wire model, the equation can be solved using the ode-45 built-in algorithm in Matlab.
The coupling coefficient, the material damping coefficient, and the initial conditions Ei (0) = Ei,0 and
dEi (0) were obtained using the method of least squares fitting to process the transient finite
dZ = Ei,0
element (FE) simulation data shown in Section 3.2. The results are shown in Figure 4. With the
propagation of the guided wave, the wave energy in the active wire was transferred to the passive
wire. The wave energy in the two wires gradually decayed after reaching a balance, which was caused
by material damping.
332
Appl. Sci. 2019, 9, 1028
Figure 4. A model of the wave energy transfer between two adjacent wires. The results of the active
wire are plotted by a red line, and the results of the passive wire are plotted by a blue line. The transient
finite element (FE) simulation data of the two wires obtained in Section 3.2. are marked by asterisks.
3. Transient Finite Element Simulations

The effective performance of the conventional transient FE technique for modeling UGW
propagation in various structures has previously been demonstrated. Here, the transient solver
ABAQUS® (Dassault System Inc., Paris, France) was employed to visually simulate UGW propagation
in two different structures: an individual wire and two contact wires. The accuracy of the results
is affected by the temporal and spatial resolution of signal acquisition in the FE simulation model.
The time step for dynamic analysis (i.e., temporal sampling) and finite element size (i.e., spatial
sampling) were determined by the maximum frequency (i.e., minimum wavelength) of the wave in
the structure, which should meet the requirements of the Nyquist–Shannon sampling theorem [31].
An optimal spatial resolution is at least eight mesh nodes per wavelength [32,33]. The temporal
sampling rate should be 10 times higher than the highest frequency component of the signal.
The parameters of the FE simulation model are summarized as Table 2.
Table 2. The parameters of the transient FE simulation model.
Diameter Length Density Young Modulus Poisson Ratio

7 mm 500 mm 7800 kg/m3 209 GPa 0.3
The dispersion of the dispersive guided wave propagating in a waveguide is relevant to the
bandwidth of the excitation signal. The wider the bandwidth of the excitation signal, the greater
the dispersion of the wave. In this paper, two types of excitation signals were used in the transient
FE simulation to investigate the dispersion performance of the guided waves in the wire. First, the
time-dependent triangular pulse signal was employed to excite a broadband signal with a frequency
bandwidth exceeding 500 kHz, as shown in Figure 5a,b. Second, the Hann-windowed, five-cycle,
60 kHz sinusoidal tone burst was designed to excite the narrow bandwidth signal with a specific center
frequency, as shown in Figure 5c,d.
333
Appl. Sci. 2019, 9, 1028

$PSOLWXGH 9
))7$ PSOLWXGH

7LPHPV )UHTXHQF\N+]

(a) (b)
(c) (d)
Figure 5. The excitation signal and fast Fourier transform (FFT) spectrum: the triangular pulse with
a broadband of 500 kHz in the (a) time domain and (b) frequency domain; the Hann-windowed,
five-cycle, 60 kHz sinusoidal tone burst in the (c) time domain signal and (d) frequency domain.
The category of the UGW mode generation in a waveguide is related to the direction of the exciting
loads. In order to achieve optimal excitation performance, the direction of vibration displacement of the
loading was set to be consistent with the wave structure of the wave mode to be excited. According to
the analysis of the wave structures of longitudinal and flexural guided waves in Section 2.1, the oblique
imposed force with uniform components in the three directions was chosen to excite longitudinal and
flexural mode waves. The perpendicular imposed force with a single component along the z-direction
was chosen to excite pure longitudinal mode waves, as shown in Figure 6.
Fx(t)=Fy(t)=Fz(t) Fz(t) Fz(t)
N+] ; ; ;

%URDGEDQG
= N+] = =
N+]
< %URDGEDQG < <
D E F
Figure 6. Guided waves in a single wire were excited using three loading methods: (a) an oblique
triangular pulse with three equal components along the x, y, and z directions, (b) a perpendicular
triangular pulse along the z-direction, and (c) a perpendicular sinusoidal tone burst along the
z-direction.
334
Appl. Sci. 2019, 9, 1028
3.1. Single Wire

In order to analyze multimodal and dispersive guided wave signals, discrete wave signals in time
and space were recorded at the receiving nodes of the FE model. Signals were mapped to dispersion
data in the frequency and wavenumber domains applying the two-dimensional (2D) fast Fourier
transform (FFT) method [34]. The essence of the 2D FFT is to perform two fast Fourier transforms on
the signal, temporally and spatially. A guided wave signal is harmonical in time and space from the
perspective of wave propagation. Therefore, the time and space domain signals f (z, t) can be mapped
to angular frequency and wavenumber domain signals F (k, ω ) as:
' ∞ ' ∞
F (k, ω ) = f (z, t)e−i(kz+ωt) dzdt (3)
−∞ −∞
c ph = ω/k (4)
where f (z, t) is the acceleration signal matrix recorded at the receiving nodes along the z axis of the
wire, c ph is the phase velocity, ω is the angular frequency, and k is the wavenumber. In this case, the
accelerations of 500 series with a spatial interval of 1 mm excited by a triangular pulse were used to
extract the 2D FFT spectrum, as shown in Figure 7.
(a) (b)
Figure 7. The phase velocity dispersion spectra obtained from nodal accelerations with different
excitation methods: (a) an oblique triangular pulse signal with three equal components, and (b) a
perpendicular triangular pulse signal with a z-axial component. The theoretical phase velocity
dispersion curves obtained in Figure 1a are represented by the black dotted lines.
The theoretical phase velocity dispersion curves overlaid on the 2D FFT spectra in Figure 7 agreed
with the transient simulation results for the longitudinal and flexural modes. As a result of the radial
vibration displacement component ur and the z-axial vibration displacement component uz , excited by
the oblique imposed forces along the x, y, and z directions, all longitudinal (i.e., L(0,1)) and flexural
modes (i.e., F(1,1), F(1,2), and F(1,3)) below 500 kHz were generated in the single wire. On the other,
the only the wave mode L(0,1) was excited by using the perpendicular triangular pulse along the
z-direction.
In order to minimize the dispersion of guided waves during propagation in the structure,
signals with a narrow bandwidth are commonly employed to excite waves for defect detection
in structures. The tone burst consisting of several cycles of Hann-windowed sine waves is a typical
narrow bandwidth signal, with the feature of: the larger the number of wave cycles, the narrower the
bandwidth of the signal. Figure 8 shows the 2D FFT spectrum extracted from the nodal accelerations
data recorded using the single wire FE model. The guided wave in the wire was excited with
a Hann-windowed, five-cycle, 60 kHz sinusoidal tone burst perpendicular to the wire end face.
335
Appl. Sci. 2019, 9, 1028
The dispersion nephogram shows a section of the L(0,1) mode with bandwidth of about 40 kHz, which
is the same as that of the excitation signal (see Figure 5b).
Figure 8. The phase velocity dispersion spectrum from nodal accelerations with a Hann-windowed,
five-cycle, 60 kHz sinusoidal tone burst along the z-direction. The theoretical phase velocity dispersion
curve of the L(0,1) mode obtained in Figure 1a is represented by a black dotted line.
3.2. Two Contact Wires

When the two ends of the multi-wire cable are subjected to an axial tension, the wire-to-wire
contact stress will appear between the adjacent wires. The magnitudes of the contact stresses are
proportional to the axial tension. A simplified wire-to-wire contact model was employed to study the
effect of contact stresses on UGW propagation in a multi-wire cable. The FE model was made up of
two 7-mm-diameter parallel straight steel wires with the same length of 500 mm. One of the wires was
active and the other was passive. This pair of wires was mechanically coupled in contact using uniform
stresses. The contact stresses were provided by a static radial pressure equaling 70% of the ultimate
tensile strength (UTS) of the wire. These contact stresses are called Hertzian contact stresses [35].
The contact condition between the two wires was modeled using a penalty friction formulation with
the friction coefficient of 0.6.
The loading conditions in the FE simulations involved two stages: static load and dynamic load.
The linear ramp signals were uniformly preloaded into two wires along the wire axes with opposite
radial directions of the two wires. The static wire-to-wire contact and friction stresses were generated
in this load step. In the second step, guided waves were excited by the dynamic load. An investigation
of the wave propagation was conducted in two cases. In Case 1, the active wire was excited with an
obliquely imposed triangular pulse force on one end face with equal components (Fx = Fy = Fz ) along
three directions. In Case 2, the active wire was excited with a Hann-windowed, five-cycle, 60 kHz
sinusoidal tone burst perpendicular to the end face along the z-direction.
Acceleration data were recorded at the surface nodes of the FE elements of the two wires along
the z-axis. The interval of the recorded points was set as 1 mm. The 2D FFT algorithm proposed in
Section 3.1 was used again to compare the two cases to observe the change in the dispersion properties
of UGWs in a pair of mechanically coupled wires. The 2D FFT transform spectra for the two cases are
shown in Figures 9 and 10.
In Figure 9, the dispersion nephograms in the 2D FFT spectra coincide with the theoretical
phase velocity dispersion curves (see Figure 1a), which is consistent with the results obtained in the
previously presented single wire case. The longitudinal mode (i.e., L(0,1)) and all flexural modes (i.e.,
F(1,1), F(1,2), and F(1,3)) below 500 kHz were generated in the active and passive wires. Guided waves
in the passive wire are transferred by contact and friction stresses between two wires. The acoustic
energy can be transferred into the adjacent wire when sufficient contact stresses occur on the interface
of two wires. As a result of the leakage of energy, the guided wave modes and energy eventually trend
to be the same, with the wave propagation in the two-wire system.
336
Appl. Sci. 2019, 9, 1028
(a) (b)
Figure 9. The phase velocity dispersion spectra of two wires from nodal accelerations with an equal
components triangular pulse force along the x, y, and z directions: (a) the active wire and (b) the passive
wire. The theoretical phase velocity dispersion curves obtained in Figure 1a are represented by the
black dotted lines.
(a) (b)
(c) (d)
Figure 10. The phase velocity dispersion spectra of wires from nodal accelerations with a
Hann-windowed, five-cycle, 60 kHz sinusoidal tone burst along the z-axial direction: (a) the active wire
and (b) the passive wire. The theoretical dispersion curve of L(0,1) obtained in Figure 1a is represented
by white dotted lines. (c) The FFT spectra of a representative received signal of the active wire and
(d) part of the representative acceleration time-domain-received signals in the active wire.
The CAN model described in Section 2.2. and the energy transfer model described in Section 2.3.
provide two different perspectives and methods to analyze the dispersion properties and propagation
behaviors of the guided wave in mechanically coupled structures. Acceleration data recorded on two
337
Appl. Sci. 2019, 9, 1028
wire surfaces with a narrow bandwidth excitation signal loaded on the active wire end face were
transformed into 2D FFT spectra, as shown in Figure 10. Dispersion nephograms in the spectra of both
the active and passive wire agree with the theoretical phase velocity dispersion curve of the single
L(0,1) mode below 500 kHz. The directions of the excitation loads determine the generation of guided
wave mode propagation in the structure. As a result of the contact stresses between the two wires, the
CAN phenomenon of the active wire was observed. The additional frequency component of 120 kHz
that doubles the excitation frequency of 60 kHz appeared in the spectrum. The frequency components
in the passive wire covered a bandwidth range of 60–120 kHz. This may have occurred because the
dispersion of guided waves in the passive wire transferred by contact stresses was more severe than
that generated by direct excitation. From the perspective of energy transfer, the guided wave energy
distribution in the active and passive wires fit the results of the energy-based model described in
Section 2.3, as shown in Figure 4.
4. Experimental Analysis

Experimental verification of the proposed methods was performed on three types of samples.
The first sample was a single 7-mm-diameter steel wire: the main component of the multi-wire bridge
cable. The second sample was a pair of mechanically coupled 7-mm-diameter steel wires in contact.
The third sample was an LGJ-400/35 ACSR cable (Tianhong Electric Power Fitting Co. Ltd., Zhejiang,
China) comprised of two layers of seven steel wires and three layers of 48 aluminum wires, which
were arranged helically in five layers (1–6–10–16–22). The diameter of the individual steel wire was
2.5 mm and the diameter of the aluminum wire was 3.2 mm. All these samples were the same length:
1.6 m.
A piezoelectric (PZT) UGW experimental inspection system was employed, as shown in Figure 11a.
The pitch-catch method was used with a broadband excitation PZT transducer pasted to one end
of the cable and the same reception transducer pasted to another cable end. The coupling radius
of the transducer was 10 mm and the height of the transducer was 22 mm, as shown in Figure 11b.
It had a broadband characteristic (frequency range of 30–150 kHz). The directivity of the transducer
was perpendicular to the contact surface. Epoxy resin glue was used to paste the transducer to the
wire surface.
Figure 11. The experimental setup for guided wave excitation and reception in three types of samples.
(a) The schematic of the experiment. Photos of (b) the piezoelectric (PZT) transducer, (c) the single wire
experiment case, and (d) the two contact wires.
338
Appl. Sci. 2019, 9, 1028
For excitation, the transducer was driven by a personal computer (PC)-controlled preamplifier
(RAM-5000, Ritec Inc., Warwick, RI, USA) with a Hann-windowed, five-cycle, sinusoidal tone burst.
The excitation signal was applied to the transducer with an amplitude of 100 Vpp (peak-to-peak value).
The detected voltage signals were filtered by a bandpass filter with a bandwidth of 150 kHz and a
center frequency of 80 kHz, and amplified by about 41 dB. Then, the signal acquisition was conducted
to the filtered signals with a sampling frequency of 2000 kHz. This routine was repeated 50 times with a
pulse repetition frequency (PRF) of 20 Hz to remove random noise and achieve a higher signal-to-noise
ratio (SNR). The signal series were averaged and processed using Matlab.
4.2. Single Wire

In these single wire experiments, four pitch-catch configuration cases were employed for guided
wave dispersion verification. The center frequency of the exciting tone burst was 60 kHz. In Case 1,
the exciting transducer was installed on the end face of the left end, and the receiving transducer was
similarly installed on the end face of the right end. In Case 2, the exciting transducer was installed
on the end face of the left end, but the receiving transducer was installed on the lateral face of the
right end. In Case 3, the exciting transducer was installed on the lateral face of the left end, but the
receiving transducer was installed on the end face of the right end. In Case 4, the exciting transducer
was installed on the lateral face of the left end, and the receiving transducer was similarly installed on
the lateral face of the right end.
The representation of a signal in the time and frequency domain can be expressed using a
spectrogram. A spectrogram is a time-frequency plot obtained using time-frequency analysis, such as
a continuous wavelet transform (CWT). It is a method used to measure experimental group velocity
dispersion curves. Figure 12 shows the time-frequency spectrum of the received signals in the four
experimental cases using the CWT method, over which the theoretical time-frequency group velocity
dispersion curves are overlaid. The theoretical propagation time was calculated using:
t = L/v group (5)
where L = 1.6 m is the length of the wire and v group is the group velocities of all existent modes at
each frequency between 0 kHz and 100 kHz. This process is equivalent to mapping the group velocity
dispersion curves from the velocity-frequency domain to the time-frequency domain.
(a) (b) (c) (d)
Figure 12. The time-frequency analysis spectrum using a continuous wavelet transform (CWT) for
a single 7-mm-diameter steel wire at a 60-kHz frequency: (a) horizontal excitation and horizontal
reception, (b) horizontal excitation but lateral reception, (c) lateral excitation but horizontal reception,
and (d) lateral excitation and lateral reception. The theoretical group velocity dispersion curves
obtained in Figure 1a are represented by the dotted line.
There were three guided wave modes in the 7-mm-diameter steel wire below 100 kHz, whose
dispersion curves were described in Section 2.1. According to the wave structures plotted in Figure 2d,e,
the radial and z-axial vibration displacements were the main components of the L(0,1) and F(1,1) modes
at 60 kHz, respectively. Z-axial wave components were generated through horizontal excitation of
the transducer. Radial wave components were generated through lateral excitation. Similarly, the
339
Appl. Sci. 2019, 9, 1028
horizontal and lateral reception methods of the transducer effortlessly received the z-axial and radial
wave components.
Thus, the L(0,1) mode was highlighted in the measured group velocity dispersion nephograms
shown in Figure 12a of Case 1. The same applied to the L(0,1) and F(1,1) modes in Cases 2 and 3
simultaneously, and F(1,1) in Case 4 individually. No torsional mode was highlighted in any of the four
cases, since neither of the two excitation methods could generate a rotational displacement component.
4.3. Two Contact Wires

In order to analyze the CAN phenomena in a multi-wire cable, a wire-to-wire cable was first
considered. Two identical 7-mm-diameter steel wires with a length of 1.6 m, used in previous
experiments, were tied together tightly and compactly using nylon cable ties along the z-axial direction,
as shown in Figure 11d. The wire that was excited to generate a UGW was labeled the active wire, and
another one was labeled the passive wire. The installation of exciting and receiving transducers in Case
2 of Section 4.2. was employed to simultaneously receive L(0,1) and F(1,1) guided waves. The center
excitation frequency of the tone burst was 60 kHz. The reception signals of the two wires were recorded
and are plotted in Figure 13. The FFT spectra of the two signals were obtained. The theoretical group
velocity dispersion curves calculated in Section 2.1. were used to determine the wave modes presented
in the signals of two wires. There were two wave packets in the time domain signals of the active
and passive wire: the direct waves of L(0,1) and F(1,1). The measured group velocities of two modes
calculated from the time-of-flight (ToF) of the wave peak in two wave packets are consistent with the
theoretical group velocity dispersion curves at 60 kHz.

60 kHz
L(0,1) direct wave
$PSOLWXGH 1RUPDOL]HG

$PSOLWXGH 9

120 kHz

F(1,1) direct wave

7LPHPV )UHTXHQF\N+]
(a) (b)

60 kHz

L(0,1) direct wave
$PSOLWXGH 1RUPDOL]HG

$PSOLWXGH 9

120 kHz

F(1,1) direct wave

7LPHPV )UHTXHQF\N+]
(c) (d)
Figure 13. The wave signals and FFT spectrum for the two contact wires with an exciting center
frequency of 60 kHz: (a) the time domain signal of the active wire, (b) the FFT spectrum of the active
wire, (c) the time domain signal of the passive wire, and (d) the FFT spectrum of the passive wire.
The appearance of the frequency components of 120 kHz in the FFT spectrum of the active and
passive wires, which was double the center frequency of the exciting signals, verifies the CAN effect of
340
Appl. Sci. 2019, 9, 1028
the two contact wires described in Section 2.2. and is consistent with the results of the transient FE
simulation in Section 3.2.
4.4. Overhead Transmission Line

As described in Section 4.1, the third experimental sample LGJ-400/500 ACSR cable consisted
of five layers (1–6–10–16–22) of wires composed of steel and aluminum. The innermost two layers
were 7.5-mm-diameter seven-wire steel strands (1–6). Since the diameter of the single steel wire
(D = 2.5 mm) was much smaller than that of the transducer (D = 10 mm), the entire steel strands
were treated as one layer. The exciting transducer was installed on the end face of the strands’ end.
The receiving transducers were installed on the lateral faces of four layers (7–10–16–22) of another
cable end, as shown in Figure 14a.

$PS 9 $P S9
VW
/D\HU VW
/D\HU
QG
/D\HU
UG
/D\HU
QG
1RUPDOL]HG (QHUJ\
/D\HU WK
/D\HU

$PS 9 $PS9
UG
/D\HU

WK

/D\HU

'LVWDQFHWR$FWXDWRUP
7LPHPV
(b) (c)
Figure 14. Guided wave propagation experiments on the aluminum conductor steel-reinforced (ACSR)
cable: (a) a photo of the exciting and receiving transducers installation, (b) the received time domain
signals of each layer, and (c) the simulated and measured time average wave energy distribution in
each layer.
Precise transient FE simulations that consider contact and friction stresses are computationally
expensive. Applications of the FE method are limited to the multi-wire structures with short lengths
and a smaller number of wires. For the sake of simplifying the description of UGW propagation
behavior in multi-wire structures, such as bridge cables and ACSR cables, the energy-based model
was used to investigate the experimental data. Guided wave signals recorded by transducers installed
on each layer of the cable end (Z = 1.6 m) were processed by Matlab and are plotted in Figure 14b.
In Figure 14b, the signals from the top to bottom were received by the transducers installed on
the first, second, third, and fourth layers of the cable (from the inside to the outside). The wave packet
marked by a circle in each signal was the direct wave excited by the exciting transducer. As a result
of the contact and friction stresses between the adjacent wires in the same and adjacent layer, wave
energy was transferred to each wire of the cable with the guided wave propagation. Each wire in
341
Appl. Sci. 2019, 9, 1028
the inner layers of the cable was in contact with four adjacent wires: left wire, right wire, wire in the
upper layer, and wire in the lower layer. Thus, the wave energy of each wire can be derived from
Equation (1). The initial conditions in the equation were determined using a least squares fit with
recorded signals. Since the active wire (first layer) was the steel strands in the center of the cable, the
normalized wave energy of the wires in the same layer was equal (second, third, and fourth layers),
as shown in Figure 14c. The measured wave energy can be obtained by integrating the received signal
over the time domain: ' t2
E(z) = Ur2 (z, t)dt (6)
t1
where t1 and t2 correspond to the time required by the fastest and slowest modes at the frequency of
60 kHz propagating from z = 0 to z = L (i.e., t1 = L/v group, f ast and t2 = L/v group,slow ), respectively,
and Ur (z, t) is the radial vibration displacement at position z. The fastest and slowest mode at 60 kHz
frequency in the 3.2-mm-diameter aluminum wire group velocity dispersion curves shown in Figure 1b
were the L(0,1) mode and the F(1,1) mode, respectively.
5. Conclusions and Outlook

In this study, an initial theoretical analysis of UGWs in a two-contact-wire system and an ACSR
cable was completed from various perspectives. A time-transient FE numerical simulation and
experimental study were conducted to validate the dispersion and wave structure properties, the
contact acoustic nonlinearity effect, and the energy transfer model of multilayer wires. The main
findings are summarized as follows:
(1) For the wave structures of the fundamental order longitudinal and flexural modes (i.e., L(0,1)
and F(1,1)) in a 7-mm-diameter steel wire, the radial vibration displacements decreased but the
axial vibration displacements increased with increasing radius. This trend became apparent with
the increase in guided wave frequency below 200 kHz. The severity of dispersion and number
of modes in a cylindrical structure were related to the bandwidth of excitation signals and the
direction of excitation loads, respectively.
(2) Guided waves with frequency f were excited to one of the two mechanically coupled wires.
Received waves with frequency 2f were observed in both wires due to the CAN effect at the
contact interface. Numerical and experimental verification were performed using a monokinetic
Hann-windowed sinusoidal excitation. There are many different wave modes simultaneously
propagating in a multi-wire structure with different velocities at different frequencies, which
could be predicted by dispersion curves. However, it is ineffective for dispersion curves to
analyze the frequency behaviors of the wave signals of a certain mode, which could be analyzed
using the CAN effect.
(3) A minimally computationally expensive energy transfer model was used to predict the wave
energy distribution in the wires located in different layers of the multi-wire cable. The approach
was applied to analyze a pair of mechanically coupled wires using transient FE simulation.
The acceleration data of the two wires were fitted to the model. The same investigation was
experimentally performed on a 1.6-m-long 55-wire ACSR cable, and the wave energy of each
layer was found to be consistent with the numerical analysis. Uniform distribution of the wave
energy in each layer of a multi-wire cable is a basic requirement of SHM using a guided wave.
The uniformity of the UGW energy depends on the contact between adjacent layers, which can
be predicted by the model, when using piezoelectric transducers installed on the outermost layer
of the cable. The model also could be used to optimize the excitation power of transducers and
determine the effective monitoring range of SHM. In industrial applications, the results of this
study could be used to improve the detection rate of the defects of wires located in the inner
layers of the cable with low signal amplitudes, which are caused by the nonuniform wave energy
distribution in different layers of the cable.
342
Appl. Sci. 2019, 9, 1028
For future research, we plan to study the UGW propagation in multi-wire cables using more
precise contact and friction models to provide guidance for the guided wave inspection of cables, and
the relationship between cable force and the CAN effect.
Author Contributions: P.Z. and Z.T. conceived and designed the models and methods presented; P.Z. developed
the experimental software tools; F.L. and K.Y. helped in conception and experiments; Z.T. helped in the simulations;
P.Z. performed the experiments and analyzed the data; and P.Z. wrote this paper.
Funding: This research was funded by the National Natural Science Foundation of China (Grant No. 61271084
and No. U1709216), the Science and Technique Plans of Zhejiang Province (Grant No. 2017C01042), the National
Key Research and Development Program of China (Grant No. 2018YFC0809000), and the Project of Inner Mongolia
Power Group (Grant No. 20184733). And The APC was funded by the National Natural Science Foundation
of China.
References
1. Shull, P.J. Nondestructive Evaluation: Theory, Techniques, and Applications; CRC Press: Boca Raton, FL,
USA, 2016.
2. Ye, X.W.; Dong, C.Z.; Liu, T. Force Monitoring of Steel Cables Using Vision-Based Sensing Technology:
Methodology and Experimental Verification. Smart Struct. Syst. 2016, 18, 585–599. [CrossRef]
3. Zhang, J.; Guo, Z.; Jiao, T.; Wang, M. Defect Detection of Aluminum Alloy Wheels in Radiography Images
Using Adaptive Threshold and Morphological Reconstruction. Appl. Sci. 2018, 8, 2365. [CrossRef]
4. Omidi, P.; Zafar, M.; Mozaffarzadeh, M.; Hariri, A.; Haung, X.; Orooji, M.; Nasiriavanaki, M. A Novel
Dictionary-Based Image Reconstruction for Photoacoustic Computed Tomography. Appl. Sci. 2018, 8, 1570.
[CrossRef]
5. Qin, L.; Ren, H.; Dong, B.; Xing, F. Development of Technique Capable of Identifying Different Corrosion
Stages in Reinforced Concrete. Appl. Acoust. 2015, 94, 53–56. [CrossRef]
6. Shin, E.; Kang, B.; Chang, J. Real-Time Hifu Treatment Monitoring Using Pulse Inversion Ultrasonic Imaging.
7. Zhang, J.; Tan, X.; Zheng, P. Non-Destructive Detection of Wire Rope Discontinuities From Residual Magnetic
Field Images Using the Hilbert-Huang Transform and Compressed Sensing. Sensors 2017, 17, 608. [CrossRef]
[PubMed]
8. Mijarez, R.; Baltazar, A.; Rodríguez-Rodríguez, J.; Ramírez-Niño, J. Damage Detection in ACSR Cables Based
on Ultrasonic Guided Waves. DYNA 2014, 81, 226–233. [CrossRef]
9. Laguerre, L.; Treyssede, F. Non Destructive Evaluation of Seven-Wire Strands Using Ultrasonic Guided
Waves. Eur. J. Environ. Civ. Eng. 2011, 15, 487–500. [CrossRef]
10. Qian, J.; Chen, X.; Sun, L.; Yao, G.; Wang, X. Numerical and Experimental Identification of Seven Wire Strand
Tensions Using Scale Energy Entropy Spectra of Ultrasonic Guided Waves. Shock Vib. 2018, 2018, 6905073.
[CrossRef]
11. Schaal, C.; Bischoff, S.; Gaul, L. Damage Detection in Multi-Wire Cables Using Guided Ultrasonic Waves.
Struct. Health Monit. Int. J. 2016, 15, 279–288. [CrossRef]
12. Zhang, P.; Tang, Z.; Duan, Y.; Yun, C.B.; Lv, F. Ultrasonic Guided Wave Approach Incorporating Safe for
Detecting Wire Breakage in Bridge Cable. Smart Struct. Syst. 2018, 22, 481–493. [CrossRef]
13. Rizzo, P. Ultrasonic Wave Propagation in Progressively Loaded Multi-Wire Strands. Exp. Mech. 2006,
46, 297–306. [CrossRef]
14. Hernandez Crespo, B.; Courtney, C.; Engineer, B. Calculation of Guided Wave Dispersion Characteristics
Using a Three-Transducer Measurement System. Appl. Sci. 2018, 8, 1253. [CrossRef]
15. Raišutis, R.; Kažys, R.; Mažeika, L.; Žukauskas, E.; Samaitis, V.; Jankauskas, A. Ultrasonic Guided
Wave-Based Testing Technique for Inspection of Multi-Wire Rope Structures. NDT E Int. 2014, 62, 40–49.
[CrossRef]
16. Raišutis, R.; Kažys, R.; Mažeika, L.; Samaitis, V.; Žukauskas, E. Propagation of Ultrasonic Guided Waves in
Composite Multi-Wire Ropes. Materials 2016, 9, 451. [CrossRef] [PubMed]
17. Schaal, C.; Bischoff, S.; Gaul, L. Energy-Based Models for Guided Ultrasonic Wave Propagation in Multi-Wire
Cables. Int. J. Solids Struct. 2015, 64–65, 22–29. [CrossRef]
343
Appl. Sci. 2019, 9, 1028
18. Haag, T.; Beadle, B.M.; Sprenger, H.; Gaul, L. Wave-Based Defect Detection and Interwire Friction Modeling
for Overhead Transmission Lines. Arch. Appl. Mech. 2009, 79, 517–528. [CrossRef]
19. Hayashi, T.; Tamayama, C.; Murase, M. Wave Structure Analysis of Guided Waves in a Bar with an Arbitrary
Cross-Section. Ultrasonics 2006, 44, 17–24. [CrossRef] [PubMed]
20. Ibáñez, F.; Baltazar, A.; Mijarez, R. Detection of Damage in Multiwire Cables Based On Wavelet Entropy
Evolution. Smart Mater. Struct. 2015, 24, 085036. [CrossRef]
21. Treyssède, F. Elastic Waves in Helical Waveguides. Wave Motion 2008, 45, 457–470. [CrossRef]
22. Li, C.; Han, Q.; Liu, Y.; Liu, X.; Wu, B. Investigation of Wave Propagation in Double Cylindrical Rods
Considering the Effect of Prestress. J. Sound Vib. 2015, 353, 164–180. [CrossRef]
23. Chillara, V.K.; Lissenden, C.J. Analysis of Second Harmonic Guided Waves in Pipes Using a Large
Radius Asymptotic Approximation for Axis-Symmetric Longitudinal Modes. Ultrasonics 2013, 53, 862–869.
[CrossRef] [PubMed]
24. Jhang, K. Nonlinear Ultrasonic Techniques for Nondestructive Assessment of Micro Damage in Material: A
Review. Int. J. Precis. Eng. Manuf. 2009, 10, 123–135. [CrossRef]
25. Legg, M.; Yücel, M.K.; Kappatos, V.; Selcuk, C.; Gan, T. Increased Range of Ultrasonic Guided Wave Testing of
Overhead Transmission Line Cables Using Dispersion Compensation. Ultrasonics 2015, 62, 35–45. [CrossRef]
[PubMed]
26. Gazis, D.C. Three-Dimensional Investigation of the Propagation of Waves in Hollow Circular Cylinders. I.
Analytical Foundation. J. Acoust. Soc. Am. 1959, 31, 568–573. [CrossRef]
27. Ditri, J.J.; Rose, J.L. Excitation of Guided Elastic Wave Modes in Hollow Cylinders by Applied Surface
Tractions. J. Appl. Phys. 1992, 72, 2589–2597. [CrossRef]
28. Seco, F.; Jiménez, A.R. Modelling the Generation and Propagation of Ultrasonic Signals in Cylindrical
Waveguides. Ultrason. Waves 2012. [CrossRef]
29. Seco, F.; Martín, J.M.; Jiménez, A.; Pons, J.L.; Calderón, L.; Ceres, R. Pcdisp: A Tool for the Simulation of
Wave Propagation in Cylindrical Waveguides. In Proceedings of the 9th International Congress on Sound
and Vibration, Orlando, FL, USA, 8–11 July 2002.
30. Biwa, S.; Nakajima, S.; Ohno, N. On the Acoustic Nonlinearity of Solid-Solid Contact with
Pressure-Dependent Interface Stiffness. J. Appl. Mech. 2004, 71, 508–515. [CrossRef]
31. Abdul, J. The Shannon Sampling Theorem—Its Various Extensions and Applications: A tutorial review.
Proc. IEEE 1977, 65, 1565–1596.
32. Datta, D.; Kishore, N.N. Features of Ultrasonic Wave Propagation to Identify Defects in Composite Materials
Modelled by Finite Element Method. NDT E Int. 1996, 29, 213–223. [CrossRef]
33. Kirby, R. Modeling Sound Propagation in Acoustic Waveguides Using a Hybrid Numerical Method. J. Acoust.
Soc. Am. 2008, 124, 1930–1940. [CrossRef]
34. Alleyne, D.; Cawley, P. A Two-Dimensional Fourier Transform Method for the Measurement of Propagating
Multimode Signals. J. Acoust. Soc. Am. 1991, 89, 1159–1168. [CrossRef]
35. Johnson, K.L. Contact Mechanics; Cambridge University Press: Cambridge, UK, 1987.
344
Article
Compensation for Group Velocity of Polychromatic
Wave Measurement in Dispersive Medium
Seung Jin Chang and Seung-Il Moon *
Department of Electrical and Computer Engineering, Seoul National University, Seoul 151-744, Korea;
jpromo8@gmail.com
* Correspondence: moonsi@plaza.snu.ac.kr; Tel.: +82-2-880-7257
Received: 21 November 2017; Accepted: 15 December 2017; Published: 18 December 2017
Abstract: The estimation of instantaneous frequency (IF) method is introduced to compensate for
the group velocity of electromagnetic wave in dispersive medium. The location of the reflected
signal can be obtained by using the time-frequency cross correlation (TFCC), following which it is
used to extract the transmitted signal from the total signal acquired. The signal propagated in the
dispersive medium is attenuated and distorted by the attenuation characteristics, which depend on
the frequency of the medium. By using the IF curve calculated for the transmitted signal, the changed
center frequency and time terms can be obtained. The obtained terms are used to compensate for
the group velocity error induced by signal distortion and attenuation. Through experiments and
simulation, the accuracy of the proposed method is 2% higher than that of the conventional method
when the signal propagates over a long distance.
Keywords: electromagnetic wave; group velocity; time-frequency domain reflectometry; dispersive medium
1. Introduction
It is important to accurately measure the group velocity of electromagnetic wave in a dispersive
medium to detect defects and measure distances. Since the group velocity is dependent on the
frequency, and dispersion occurs when a polychromatic wave composed of multiple frequencies
propagates in a dispersive medium, it is difficult to measure the exact group velocity [1]. group velocity
correction is necessary when a signal propagates in any medium other than air. Detection methods
based on the electromagnetic wave, called reflectometry methods, are useful for fault localization
and monitoring of the health of a cable with insulator. Reflectometry methods can be categorized
into time domain reflectometry (TDR), frequency domain reflectometry (FDR), and time-frequency
domain reflectometry (TFDR) [2–7] depending on the incident signal type. TDR and FDR use a
step pulse and sinusoidal pulse defined in the time domain and frequency domain, respectively.
They have been widely used in cable diagnostics owing to their ease of implementation. However,
since the incident signals of TDR and FDR defined one domain are used, it is difficult to distinguish
the acquired signal from the noise. In contrast to TDR and FDR, TFDR has a higher signal-to-noise
(SNR) because the incident signal of TFDR is used as a chirp signal that has characteristics in both the
time and frequency domains. In TFDR, the impedance discontinuity distance is measured by obtaining
the time delay of the reflected signal generated from the impedance discontinuity point for a given
group velocity. Since the group velocity is determined by the permittivity and permeability of the
propagation medium, the propagation can be obtained from the information about the propagation
medium. The time-frequency cross correlation function (TFCC) which is a measure of the similarity of
two signals, and calculates the time delay, is based on the Wigner-Ville distribution, which is an energy
distribution in time-frequency analysis. However, since the group velocity depends on the frequency,
if the bandwidth is wide or distortion is increased by long-distance propagation, the error of the group

Appl. Sci. 2017, 7, 1306
velocity also becomes large. Especially, in case of submarine cable lengths of several hundred km,
speed compensation is essential. Our research group has been investigated the compensation for group
velocity [8]. The compensation method proposed in the previous method [8] does not consider the
sweep rate of the reflected signal to be changed. That is, since the frequency components included
in the chirp signal have different velocities, compensation is performed without taking into account
dispersion, which causes an measurement error.
In this paper, we propose a method for compensating the group velocity of poly-chromatic wave
in a dispersive medium based on instantaneous frequency (IF) estimation. Using TFCC, the location
of transmitted signal is obtained and the signal is extracted from the total acquired signal. IF curves
of incident and transmitted signal are derived by using the phase unwrapping process. The shifted
center frequency and shortened time duration of transmitted signal are obtained based on the Fast
Fourier Transform (FFT) and estimation of the IF curve. The group velocity compensation is carried
out through the derived terms.
2. Group Velocity Measurement Based on Electromagnetic Wave Theory

The proposed group velocity measurement method consists of two main parts. The first part
comprises the transmitted signal detection method using TFCC. After the first step, the time offset
and shifted center frequency of the transmitted signal are used to compensate for the group velocity.
In electromagnetic theory, the group velocity of an electromagnetic plane wave can be derived as
follows [1]:
c c
vp( f ) = 0 = μ s −1
, (1)
(( f ) · μ( f )) s −∞ 1/2
( ∞ + f ) · (1 + f )1/2
1+ j f (1+ j f m )2
r
where f , f r , f m are the operating frequency, relaxation frequency at which the imaginary part of
the complex permittivity reaches a maximum, and resonant frequency, respectively. ∞ , s are the
permittivity at infinity frequency and static permittivity, respectively. μs is the static permeability.
According to (1), the group velocity depends on the operating frequency. When the polychromatic
wave comprising several frequencies propagates in the dispersive media, the wave is distorted and a
group velocity error occurs because of the frequency dependency of group velocity. Especially when
the signal is transmitted over a long distance, the more high frequency components are attenuated,
and velocity error is induced. In other words, as the signal propagates, the group velocity changes
continuously with the propagation distance. In this paper, the incident signal is the signal that we
injected into the cable, the reflected signal is generated at the cable termination, and transmitted
signals are signals that are acquired by the oscilloscope through the inductive couplers as the incident
signal flows along the cable, and the signals acquired after being reflected at the cable termination are
classified as reflected signals, and the reflected signal is included in the transmitted signal. We use the
Gaussian linear chirp signal as the incident polychromatic wave, and the incident signal is represented
as follows:

Ae−αt /2+ j(0.5ξ 1 t +ωt) , t= 0≤ t ≤ T1
2 2
s(t) = (2)
0, otherwise.
where T1 is the duration of the incident signal, A is the amplitude, ξ 1 is the normalized angular
frequency sweep rate, and ω is the normalized angular frequency. The transmitted signal is expressed
as follows:

η · Ae−αt /2+ j(0.5ξ 2 (t−d) +ω (t−d)+φ) , d≤t≤d+T2
2 2
r (t − d) = (3)
0, otherwise.
346
Appl. Sci. 2017, 7, 1306
where η is the magnitude of the attenuation coefficient at the travelling distance, and d is the time
delay of the transmitted signal. Also, because the transmitted signal has high frequency attenuation,
we define the changed parameters ξ 2 and T2 as frequency sweep rate and time duration of transmitted
signal. The normalized cross-correlation between the incident signal and the transmitted signal is used
to detect the transmitted signal from the cable termination. This can be expressed as follows [3]:
N −1 1
sk rk−t
Csr (t) = ∑ √
Es Er
(4)
k =0
where Es is the energy of the incident signal in Wigner-Ville distribution, Er is the energy of the
1
transmitted signal, and is the correlation operator. Through the normalized cross-correlation
process, the transmitted signal can be extracted from the total acquired signal.
3. Compensation of Group Velocity in Dispersive Medium

For the compensation of group velocity, we measure the IF curves of incident and transmitted
signals. To acquire IF curves, the incident and transmitted signals are transformed through Hilbert
transform for as following analytic representation:
z(t) = r (t) + jH t {r } = Me jΦ (5)
where, Ht { x } is a output at a time t of Hilbert transform filter applied the signal r, M and Φ are the
magnitude and the instantaneous phase of the transmitted signal. And then, the instantaneous phase
is derived as following:
abs( j ∗ Hilbert(r (t)))

ξ 2 (t − d)2 + ω (t − d) + φ = tan−1 (6)
abs( Hilbert(r (t)))
where, abs(·) means the absolute function. Figure 1 shows the illustration of the compensation of
IF curves. In Figure 1, f s,c and f r,c are the center frequency of incident signal and transmitted signal.
To derive the time shifted terms, a( f ), b( f ), of transmitted signal according to operation frequency,
we obtained the shifted term using the delayed incident signal by the time delay, d, and the reflection
signal as follows.
f r,c − f 0 f r,c − f 0 f s,c − f r,c

a( f ) = , b( f ) = , c( f ) = (7)
ξ1 ξ2 ξ1
The concept of group velocity of polychromatic signal is associated with narrow-band pulses
concentrated in the neighborhood of center frequency, ω0 , with an effective frequency band
|ω − ω0 | ≤ Δω, where Δω
ω0 . Furthermore, the compensated time delay between incident signal
and transmitted signal is derived as following:
tcom = d + (b( f ) − a( f )) − c( f ) (8)
And then, compensated group velocity is derived as follows: vcom = l/tcom . where, l is the known
length of target cable. In future work, we derive the frequency and group velocity, v( f ), curve based
on (1). And then, we assume that the center frequency of transmitted signal is linearly decrease with
time. According to this assumption, the center frequency of transmitted signal can be expressed as:
f r,c (t) = a · t + b where, a < 0 and b are the constant. The travelling distance, D can be derived as:
. t +t
D = t 0 com v( f r,c (t))dt. This article compensates the speed with time delay compensation, and next it
0
will develop this to estimate the distance through accurate propagation speed and time delay without
travelling distance information.
347
Appl. Sci. 2017, 7, 1306
Frequency (f ) Incident signal s(t-d)

s(t) Transmitted signal
r(t)
fs,c
fr,c a(f)
c(f)
fo
b(f)
d
Compensated time delay
tcom Time (t)
to
Figure 1. Illustration of compensation method based on IF curves.
4. Experimental Results
Figure 2 shows the illustration of the proposed system. The system is composed of (1) digital
phosphor oscilloscope (DPO); (2) arbitrary waveform generator (AWG); and (3) inductive couplers.
AWG generates the reference signal and apply the signal to target cable though the inductive coupler.
The inductive coupler uses the electromagnetic induction phenomenon to apply a signal to the
cable without connection between the cable core and signal line. In this paper, we use three inductive
couplers and coupler 1 is used to apply a signal to cable, couplers 2 and 3 acquire the signal. An incident
signal propagates along the cable and was acquired in the oscilloscope through the couplers 2 and 3.
For ease of understanding, the signals were numbered according to the order in which they were
acquired. The transmitted signals acquired through the couplers 2 and 3, before the signal being
reflected, were numbered 1 and 2 and located at 0 m and 40 m, respectively. The distance propagated
based on the first acquired signal though coupler 2 becomes the position of the signal. The signals
reflected from the cable termination were in turn acquired via couplers 3 and 2, which were located at
80 m and 120 m. In order to verify the variation of group velocity due to the dispersion, we converted
the experimental system into an equivalent circuit model using the simulation tool. Through the
tuning of loss factor in the simulation tool, we conducted the simulation to compare the signal passing
through the lossy medium with lossless medium. The comparison results are shown in Figure 3.
The results consist of 3 types simulations: (a) lossless cable: 40 m; (b) lossy cable: 40 m; (c) lossy cable:
80 m. To verify the effect of loss factor, we compared the transmitted signal in lossy medium with
that in lossless medium (simulations: (a) and (b)). Also, simulations were conducted with different
cable lengths in order to analyze whether the group velocity of the signal varies due to the reflection
(simulations: (b) and (c)). As shown in Figure 3a, the incident signals of each simulation are identical,
but the reflected signals of each did not match. Figure 3b shows an enlarged view of the reflected
signals. The highest peak in the time domain of the signal can be thought of as the highest energy
point, and the group velocity of the signal can be determined by the time delay between these peak
points, and this time delay is called the time of arrival of group velocity. Comparing the waveforms of
(a) and (b), the time delay between the peak points of the reflected signal generated at 40 m and 80 m
are 0.102 · 10−6 s and 0.207 · 10−6 s, respectively. If there was no change in group velocity depending
on the travel distance, the time of arrival of group velocity of the reflected signal generated at 80 m
348
Appl. Sci. 2017, 7, 1306
had to be 0.204 × 10−6 m. More time delays mean slower group velocity. These results show that
the group velocity is decreasing as the signal propagates. Since the incident signal is a positive chirp
signal, the rear part of the signal in time domain contains a high frequency component. As shown in
the reflected signals of red and black line of Figure 3b, in the front part, the zero-crossing points of
each signal are matched, but the zero-crossing points in the rear part do not coincide with each other.
These results indicate that the higher frequency components of the signal are attenuated as the signal
propagates in the lossy medium. Comparing the reflected signals generated at 80 m, we verified that
the reflection only affect the magnitude of signal, not group velocity. Figure 4a shows the total acquired
signal from the inductive couplers after signal restoration process [9,10]. As seen in the fourth signal in
Figure 4a, because the signal is difficult to distinguish from the noise, TFCC is used to roughly find the
position of the transmitted signal. To evaluate the accuracy of proposed method, we solved the true
value of group velocity using the time of arrival of group velocity of the signals in the time domain.
However, the reflected signal at 120 m is too small to find out due to the attenuation. Because of this,
it is very difficult to obtain the group velocity through the time delay between the highest peak points
of the signal in the time domain, and the group velocity in the farther than 120 m can not be calculated.
The TFCC graph is shown in the Figure 4b. As the signal propagates, attenuation of the signal occurs,
which slows the group velocity and increases the time delay. Figure 4b depicts a TFCC graph based on
the constant group velocity measured on 40 m. Therefore, the distance errors of 80 m and 120 m is
getting larger. Based on unwrapping algorithm and Hilbert transform, the instantaneous phase of the
transmitted signal can be derived and shown in Figure 5a. The signal having the positive slope of the
instantaneous phase was extracted, and the frequency band of the extracted signal was obtained by
FFT algorithm. In Figure 5a, the time duration of the signal is obtained by extracting the signal where
the slope of the instantaneous phase is positive. The Figure 5b shows the frequency band of second
signal of total signal. The changed values, time duration and frequency region, were obtained and
substituted into Equation (7) to compensate the group velocity. As seen in Table 1, the group velocity
in TFCC method seems to be equal regardless of the propagation distance. On the contrary, in the
proposed method, the shifted terms , a( f ), b( f ), c( f ), can be obtained from the center frequency, f rc ,
of the received signal from Figures 1 and 5. The compensation time delay term is calculated according
to Equation (8). The travelling distance, D, is the known value of the cable length and is the integral
of the velocity determined by the center frequency of the transmitted signal. Based on the derived
compensation time delay, tcom , and the distance value, D, the average velocity during propagation of
the transmitted signal can be obtained and are shown in Table 1. The measurement values in Table 1
were calculated by the time of arrival of group velocity of signal in Figure 5a. As the distance between
each signal is set as 40 m, the group velocity, measurement value, can be obtained by dividing the
distance by the measured group delay. The accuracy values in Table 1 were calculated by dividing
the group velocity of proposed method by measurement value. As seen in Table 1, when the signal
propagates to a short distance, the existing method is highly accurate, but the existing method does
not reflect the change in group velocity when the signal propagates over a long distance. On the other
hand, the proposed method compensates the group velocity change due to the dispersion, so that the
accuracy of group velocity is good regardless of the distance.
349
Appl. Sci. 2017, 7, 1306
Figure 2. Experimental setup.
Figure 3. The results of acquired signals: 40 m lossy cable, 80 m lossy cable, lossless cable (a) total
signal; (b) reflected signal.
350
Appl. Sci. 2017, 7, 1306
0.6
Magnitude [V]
0.4
0.2
0
-0.2
-0.4
0 20 40 60 80 100 120
distance [m]
0.8
TFCC value
0.6
0.4
0.2
0 20 40 60 80 100 120
distance [m]
Figure 4. The results of (a) acquired signals, (b) TFCC graph.
Table 1. Estimation results of group velocity.
Group Velocity
Cable Length: 40 m Cable Length: 80 m Cable Length: 120 m
TFCC 1.82 · 108 [m/s] 1.81 · 108 [m/s] 1.81 · 108 [m/s]
accuracy 100 [%] 99 [%] 97 [%]
proposed method 1.83 · 108 [m/s] 1.81 · 108 [m/s] 1.76 · 108 [m/s]
accuracy 99 [%] 99 [%] 99 [%]
measurement
1.82 · 108 [m/s] 1.80 · 108 [m/s] 1.77 · 108 [m/s]
value
5. Conclusions
In this paper, we proposed the group velocity compensation method using the derived time and
center frequency offset terms based on the estimation of instantaneous frequency (IF). The proposed
method can be divided two part. The one was the transmitted signal detector algorithm based on
TFCC and multiple inductive couplers system. The second is the compensation algorithm, and the
compensation terms was obtained based on the IF curve which was derived from Hilbert transform
and phase unwrap algorithm of the transmitted signal. The variation of group velocity of chirp
signal by dispersion in lossy medium was verified using a simulation tool. Through the comparison
experiments with compensation without consideration of sweep rate change and existing methods
using TFCC, superiority of the proposed method was proved. Although the group velocity of a signal
propagating a short distance has similar accuracy in both the conventional method and the proposed
method, when the signal propagates over a long distance (120 m), the proposed method has 2% better
accuracy than the conventional method. This paper proposes a new method to compensate the group
velocity error due to the dispersion of the chirp signal in the lossy medium, and this method can be
applied to the detection of defect, it is possible to localize the fault in long-distance lines, such as
submarine HVDC cable, without error.
351
Appl. Sci. 2017, 7, 1306
Figure 5. (a) Estimation of instantaneous phase, (b) frequency band of transmitted signal.
Author Contributions: Seung Jin Chang made a theory, designed the experiments and wrote the paper;
Seung-Il Moon analyzed the data.
References
1. Cho, J.; Kim, H.; Jung, K.-Y. Simple transmission line model suitable for the electromagnetic pulse coupling
analysis of twisted-wire pairs above ground. IEICE Electron. Express 2016, 13, 20160149.
2. Chowdary, K.M.; Majetich, S.A. Frequency-dependent magnetic permeability of Fe10Co90 nanocomposites.
J. Phys. D Appl. Phys. 2014, 47, 175001, doi:10.1088/0022-3727/47/17/175001.
3. Chang, S.J.; Lee, C.K.; Lee, C.-K.; Han, Y.J.; Jung, M.K.; Park, J.B.; Shin, Y.-J. Condition monitoring of
instrumentation cable splices using Kalman filtering. IEEE Trans. Instrum. Meas. 2015, 64, 3490–3499,
doi:10.1109/TIM.2015.2444260.
4. Lee, C.K.; Park, J.B.; Shin, Y.J.; Yoon, T.S. High resolution LFMCW radar system using model-based beat
frequency estimation in cable fault localization. IEICE Electron. Express 2014, 11, 20130768.
5. Lee, S.H.; Lee, C.K.; Park, J.B.; Choi, T.H. Diagnostic method for insulated power cables based on wavelet
energy. IEICE Electron. Express 2013, 10, 20130335.
6. Kwak, K.S.; Doo, S.; Lee, C.K.; Yoon, T.S. Reduction of the blind spot in the time-frequency domain
reflectometry. IEICE Trans. Electron. 2008, 5, 265–270.
7. Lee, C.K.; Kwak, K.S.; Yoon, T.S.; Park, J.B. Cable Fault Localization Using Instantaneous Frequency
Estimation in Gaussian-Enveloped Linear Chirp Reflectometry. IEEE Trans. Instrum. Meas. 2013, 62, 129–139,
doi:10.1109/TIM.2015.2444260.
8. Shin, Y.J.; Song, E.S.; Kim, J.W.; Park, J.B.; Yook, J.G.; Powers, E.J. Time-frequency domain reflectometry for
smart wiring systems. Proc. SPIE 2002, 4791, doi:10.1117/12.451709.
9. Chang, S.J.; Lee, C.K.; Shin, Y.-J.; Park, J.B. Multiple Resolution Chirp Reflectometry for Fault Localization
and Diagnosis High Voltage Cable in Automotive Electronics. Meas. Sci. Technol. 2016, 27, 125006,
doi:10.1088/0957-0233/27/12/125006.
10. Selesnick, I.W.; Figueiredo, M.A.T. Signal restoration with overcomplete wavelet transforms: comparison of
analysis and systhesis priors. Proc. SPIE 2009, 7446, doi:10.1117/12.826663.
352
applied
sciences
Article
Analysis of Guided Wave Propagation in a
Multi-Layered Structure in View of Structural
Health Monitoring †
Yevgeniya Lugovtsova 1, *, Jannis Bulling 1 , Christian Boller 2 and Jens Prager 1
1 Bundesanstalt für Materialforschung und –prüfung (BAM), 12205 Berlin, Germany;
jannis.bulling@bam.de (J.B.); jens.prager@bam.de (J.P.)
2 Chair of NDT and Quality Assurance (LZfPQ), Saarland University, 66125 Saarbruecken, Germany;
c.boller@mx.uni-saarland.de
* Correspondence: yevgeniya.lugovtsova@bam.de
† This paper is an extended version of our paper published in 9th European Workshop on Structural Health
Monitoring (EWSHM 2018), 10–13 July 2018 in Manchester, UK.
Received: 25 September 2019; Accepted: 25 October 2019; Published: 29 October 2019
Abstract: Guided waves (GW) are of great interest for non-destructive testing (NDT) and structural
health monitoring (SHM) of engineering structures such as for oil and gas pipelines, rails, aircraft
components, adhesive bonds and possibly much more. Development of a technique based on GWs
requires careful understanding obtained through modelling and analysis of wave propagation and
mode-damage interaction due to the dispersion and multimodal character of GWs. The Scaled
Boundary Finite Element Method (SBFEM) is a suitable numerical approach for this purpose
allowing calculation of dispersion curves, mode shapes and GW propagation analysis. In this article,
the SBFEM is used to analyse wave propagation in a plate consisting of an isotropic aluminium layer
bonded as a hybrid to an anisotropic carbon fibre reinforced plastics layer. This hybrid composite
corresponds to one of those considered in a Type III composite pressure vessel used for storing gases,
e.g., hydrogen in automotive and aerospace applications. The results show that most of the wave
energy can be concentrated in a certain layer depending on the mode used, and by that damage
present in this layer can be detected. The results obtained help to understand the wave propagation in
multi-layered structures and are important for further development of NDT and SHM for engineering
structures consisting of multiple layers.
Keywords: lamb waves; composite; ultrasonic testing; numerical modelling; pressure vessels
1. Introduction
Guided waves (GW) are of great interest for non-destructive testing (NDT) and structural health
monitoring (SHM) of engineering structures such as for oil and gas pipelines, rails, aircraft components,
adhesive bonds and possibly much more [1–5]. These waves can propagate over relatively long
distances as long as the structure’s cross-section stays constant and the difference of acoustic impedance
to the surrounding environment is large. It may further allow investigation of inaccessible areas of
the structure under given circumstances. To minimise complexity in GW based analysis mainly the
fundamental (i.e., first symmetric and antisymmetric) modes are used in the lower frequency range,
since those are well understood and can be excited, measured and consequently analysed without
difficulty [6,7]. Techniques based on GWs are therefore still under investigation to explore the more
complex applications such as wave propagation in hybrid structures. For instance, some structures
are composed of a fibre-reinforced composite and a metal of constant cross-section [8–13]. This is
a popular design of composite pressure vessels (COPV) used for gas storage. To date, a standard

Appl. Sci. 2019, 9, 4600
test procedure for this type of vessels does not exist and needs to be developed. Various approaches
have been proposed to assess the current degradation state of COPVs. Destructive tests of some
of the vessels running in parallel with vessels being in service were suggested in [14]. Another
approach is acoustic emission monitoring for prediction of the residual lifetime of COPVs [11,15].
Many researchers have been exploring an option of monitoring pressure vessels using integrated
optical fibres with Bragg grating sensors [16–20]. Another promising approach is guided wave-based
monitoring [10,12,13,21–23], which has several advantages when compared to alternative monitoring
techniques. One of the advantages is that the approach is non-destructive and another is the approach’s
holistic character with respect to detection and localisation of damage.
To examine different features of guided waves with respect to the limits of detectability of
damage in a structure to be monitored, specifically also with regard to location when considering the
hybrid material mentioned above, higher-order modes provide an interesting option to be explored.
These modes, when considered over the whole thickness of the laminate, do not necessarily have
smaller wavelengths but rather possess more complex mode shapes. This, in turn, may allow damage
located in different depths to be found and differentiated. Due to dispersion and the multimodal
character of GWs, development of a damage detection technique based on GWs requires careful
numerical modelling such that a realistic analysis of wave propagation and mode-damage interaction
can be performed.
At first, a tool for calculation of dispersion curves and mode shapes is required. For this purpose,
efficient approaches have been developed and applied to layered structures, e.g., the transfer matrix
method [24], the discrete layer method [25], the semi-analytical finite element method [26,27] and the
scaled boundary finite element method (SBFEM) [28]. Based on these curves, appropriate modes can
be identified and considered for further analysis. Next, a numerical tool for modelling the propagation
of the chosen GW modes is required. Numerous methods have been applied to wave propagation
problems such as finite differences [29], finite volumes [30–32], and finite element methods [11,33],
to just name a few. An extensive review of simulation methods for guided wave propagation analysis
can be found in [34]. In general, each method has its advantages and disadvantages depending on the
given problem, and there is no universal method suitable for solving every problem efficiently.
In this article, the SBFEM is used to analyse GW propagation in a plate consisting of an
isotropic aluminium layer bonded to an anisotropic laminate made of carbon fibre reinforced plastics
(CFRP). Compared to other methods, the SBFEM does not require full discretisation, thus reducing
the computational effort required significantly. It is suitable for calculation of dispersion curves,
mode shapes and modelling of the GW propagation [28,35,36]. The dispersion curves are computed
with the SBFEM by discretising the cross-section (e.g., the plate thickness) of an infinite domain
using high-order elements, as shown as an example in Figure 1a. Multi-layered structures can be
easily modelled using one element per layer having either isotropic or anisotropic material properties.
This method is also highly efficient for modelling of the wave propagation in structures which have a
constant cross-section. Because only the cross-section is discretised, the number of finite elements used
in the SBFEM reduces drastically when compared to common finite element methods, as can be seen as
a comparison between the SBFEM discretisation in Figure 1b and the FEM discretisation in Figure 1c,
respectively. Another feature of the SBFEM is that an infinite domain, a so-called unbounded domain,
as shown in Figure 1a, can be coupled with the domain of interest, thus avoiding unwanted reflections
and by that simplifying the analysis of the wave propagation.
To be able to distinguish a damage in the aluminium from a damage in the CFRP in the hybrid
composite addressed above, fundamental and higher-order GW modes have been determined and
analysed. Additionally, to some research performed and reported in [13], a delamination case has been
studied and is reported here proposing a new damage detection methodology, which allows damage
localisation within the different layers. The results show that most of the wave energy (displacement)
concentrates in a certain layer of the hybrid composite depending on the mode being considered
and with this a damage present in the respective layer may be detected. The results obtained help
354
Appl. Sci. 2019, 9, 4600
to understand the wave propagation in such a multi-layered structure which may further help GW
based techniques to be enhanced in the sense of NDT and SHM systems to be developed for hybrid
composite structures.
Y
∞ Y Y
∞
X X X
(a) (b) (c)
Figure 1. Schematic discretisation of: (a) an unbounded (infinite) domain in the scaled boundaty finite
element method (SBFEM) (b); a domain in the SBFEM; and (c) in classical FEM.
2. Methodology
Tubular composite pressure vessels are designed in a way that failure mainly happens in their
cylindrical part. There, axial (longitudinal) cracks in the metal liner, as well as fibre fracture and epoxy
matrix cracks in the CFRP overwrap, can occur [14,37,38]. These types of damage, together with an
impact damage [39,40], which may happen too, are the most critical and crucial issues with regard to a
safe use of the COPV. In the work presented here and for reasons of simplification, the end caps of
the vessel are neglected, and wave propagation is modelled in the circumferential (hoop) direction
only. This will not only allow to detect axial cracks but also to concentrate on the circumferential
CFRP plies, which are holding most of the pressure in the vessel. Furthermore, it allows damage
detectability in such a hybrid material to be studied still under less complex loading conditions
before moving to the more complex ones. In terms of the GW propagation, the cylindrical structure
is therefore approximated to a plate, being allowed if the ratio of radius to thickness is larger than
10:1 [41]. The scope of this paper is merely to investigate the GW propagation in a multi-layered plate
first while cylindrical structures need to be discussed in some future paper later. Figure 2 shows the
configuration used.
Z
X
CFRP
Aluminium
Y
Aluminium
CFRP X
Figure 2. Sketch of a simplified structure of a COPV with corresponding sensor placement, adapted
from [13].
The numerical model considered consists of a plate made of a 4 mm CFRP laminate with a
[90/0/90/0] layup (from top to bottom), followed by a 2 mm aluminium layer. Modelling was
performed under the two-dimensional plain strain assumption. The CFRP plies were modelled
separately as a stack in accordance to their lamination direction being transversally isotropic with a
1 mm thickness of each. Material properties from the ANSYS database were used in the model and are
listed in Tables 1 and 2, where ρ is density, E is Young’s modulus, G is shear modulus, and ν is Poisson’s
ratio. Due to the production process of COPVs, a firm connection between the aluminium liner and
CFRP overwrap is achieved, which allows a firm bonding between the two parts of the structure
355
Appl. Sci. 2019, 9, 4600
to be considered. To calculate the dispersion curves and mode shapes, the plate thickness has been
approximated using finite elements of the eighth order, whereas the domain in the wave propagation
direction is described analytically, as shown in Figure 1a. For the analysis of wave propagation, an
excitation domain, an evaluation domain, a damage domain and two unbounded domains at both
ends of the plate were used. The schematic of the numerical model is shown in Figure 3.
Table 1. Material properties of the epoxy carbon ply used (E1 is in the fibre direction).
ρ (kg/m3 ) E1 (GPa) E2 , E3 (GPa) G12 , G13 (GPa) ν12 , ν13 ν23

1490 121 8.6 4.7 0.27 0.4
Table 2. Material properties of aluminium used.
ρ (kg/m3 ) E (GPa) ν
2770 70 0.33
Evaluation
Excitation points
90°
1 1 1 1
90°
Delamination 90°
∞ 90° ∞
Crack
2
Al
Y
100 500 100 10 (unit: mm)
X
Figure 3. Schematic of the numerical model used in the SBFEM.
When actuating/deflecting the laminate at virtually any location over its cross-section, different
mode shapes will be generated in accordance to the frequency being applied. To minimise the effect
of dispersion by minimising a change in group velocities, some centre operation frequencies can be
chosen, which have been marked with circles in Figure 4. For those, the mode shapes have been
determined for which examples are shown in Figure 6. These mode shapes have further been used
as the excitation to simulate a respective mode within the laminate and to understand the resulting
consequences. The excitation was performed in both horizontal and vertical directions where the load
distribution over the plate has been in accordance to the shape of the mode having been considered.
These were pre-calculated using the unbounded domain in the SBFEM, for which an example is shown
in Figure 1a. Even though such an excitation can be hardly realised in real application it has been used
here for reasons of simplification and validation only and mainly to study the behaviour of a single
respective mode. The excitation in time was a Hanning-windowed tone burst with the centre frequency
of the desired mode. Data were evaluated using 4096 points over a 500 mm distance resulting in a
0.12 mm step. Modelling was performed in the frequency domain allowing the use of infinite domains.
For GW to be accepted for COPVs’ health monitoring, damage must be detected in any part of
the structure, be it the aluminium liner or the CFRP overwrap. To allow for damage detection in both
parts, modes showing characteristic interaction with damage, therefore, need to be identified. For this,
damage in the aluminium part was modelled by disconnecting two SBFEM domains, representing
a crack of 1 mm in depth and zero width, respectively. Damage in the CFRP part was modelled by
disconnecting two SBFEM boundaries, representing a delamination of 10 mm in length and zero width
between two CFRP-plies. The delamination was modelled at two different positions in two separate
simulation runs. The positions were between the second ply and the third ply (Position (1)), and the
third ply and the fourth ply (Position (2)) (counted from top to bottom). Figure 3 schematically shows
356
Appl. Sci. 2019, 9, 4600
the different locations of the damages. It further shows that the excitation has been chosen to act along
the thickness while the recordings have been made on the top and bottom of the hybrid composite as
well as between the CFRP and aluminium part respectively being denoted as “evaluation points”.
12,000
7 8
10,000 3 5 6 9 10
Phase velocity (m/s)
*URXSYHORFLW\PV

8000
4
2

6000

4000

2000 1

0
0 5 10 15
Frequency (Hz) 10 5 )UHTXHQF\+]
(a) (b)
Figure 4. Dispersion curves of: (a) phase and (b) group velocities, calculated for the infinite
aluminium-CFRP hybrid plate using SBFEM. The first ten guided wave modes are numbered in numerical
order. The circles mark modes, as chosen for numerical modelling. Recreated from original data from [13].
3. Results
Dispersion curves for the infinite aluminium-CFRP hybrid plate are presented in Figure 4.
The circles mark modes at their centre operation frequencies, as chosen for the numerical modelling.
These points were selected based on the minimal change of the group velocity, which in turn allows
dispersion effects of the mode to be minimised. All values from dispersion curves and the number
of cycles in the tone burst used for modelling are summarised in Table 3. The number of cycles was
chosen based on the frequency. The higher the frequency was, the more cycles were used in the pulse.
This allows the frequency bandwidth to be kept narrow, which also minimises the dispersion effects,
allows the excitation of undesired modes to be avoided beyond their cut-off frequencies and reduces
the computational time. The longest computational times on the standard desktop PC used for the
determination of the dispersion curves and wave propagation were 5 s and 278 s, respectively.
Table 3. Modes used for numerical modelling.
Mode Frequency Group Velocity Phase Velocity Wavelength Number of Cycles

kHz m/s m/s mm
1 100 2033 1562 15.6 10
2 475 2565 2507 5.3 20
3 950 2864 2761 2.9 30
4 400 4981 5839 14.6 20
5 475 3175 7132 15 20
6 700 4238 5431 7.8 20
7 860 5440 10,005 11.6 30
3.1. Damage in the Aluminium Liner

The signals obtained from the evaluation points, shown in Figure 3, were analysed by means of
a 2D Fast Fourier Transform (FFT). The resulting dispersion map is in the frequency–wavenumber
domain. It reveals different modes propagating in the plate and reflecting from the crack in the
aluminium liner. The frequency–wavenumber spectra were superimposed with the dispersion curves
to identify different modes, and the results are shown in Figure 5. The negative wavenumbers represent
incident waves, whereas positive wavenumbers represent reflected waves. Two cases are compared
in this figure, for the excitation of Mode 2 at 475 kHz (see Figure 5a,b) and of Mode 5 at 475 kHz
357
Appl. Sci. 2019, 9, 4600
(see Figure 5c,d), where in the cases in-plane and out-of-plane components have been determined,
respectively. These examples are shown for the evaluation points positioned at the interface between
aluminium and CFRP. Even though the excitation was performed by applying the corresponding mode
shape at the central frequency of the desired mode to be excited, the excitation of other modes could
not be avoided. However, these modes have much smaller displacements when compared to the mode
having been explicitly excited. In the case of Mode 2 excitation, Mode 4 was excited too, whereas
excitation of Mode 5 led to the excitation of Modes 1, 2 and 4, as shown in Figure 5. Because of the
same excitation frequency and the number of cycles in the pulse, but different mode shapes, different
modes were excited. The results show that, regardless of how many modes were excited in the plate,
only Modes 2 and 4 reflect from the crack in the frequency range used.
(a) (b)
(c) (d)
Figure 5. Frequency–wavenumber spectra of wave modes propagating in the aluminium-CFRP plate
and reflecting from the 1 mm crack in the aluminium: (a) for in-plane and (b) out-of-plane components
while exciting Mode 2 at 475 kHz, and (c) for in-plane and (d) out-of-plane components while exciting
Mode 5 at 475 kHz.
Considering the shapes of Modes 2 and 5, shown in Figure 6a,b, respectively, it is observed
that Mode 5 has a very small displacement in the aluminium at this frequency, compared to the
CFRP part and Mode 2. This indicates that the energy (displacement) of Mode 5 at the frequency
of 475 kHz is concentrated mainly in the CFRP layers, and thus the damage in the aluminium layer
cannot be found. The frequencies for all modes chosen for modelling, including the detectability of all
damages defined, are summarised in Table 4. As regards the crack in the aluminium liner only Modes 5
and 7 do not interact. Considering the mode shapes, Mode 5 has a very small displacement in the
aluminium layer compared to the CFRP plies (see Figure 6b). The same holds for Mode 7, not shown
here for brevity. In contrast, other modes that were investigated have comparable amplitudes in both
aluminium and CFRP parts (see Figure 6a,c,d). In Figure 7a,b, the dispersion curves of the coupled
358
Appl. Sci. 2019, 9, 4600
6 mm aluminium-CFRP plate are compared to the dispersion curves of a single 2 mm aluminium plate
and a single 4 mm CFRP plate, respectively. The dashed lines represent four guided wave modes in
the 2 mm aluminium plate for the frequency range chosen. The modes propagating in the 4 mm CFRP
plate are marked with dashed-dotted lines in Figure 7b. The combination of these two structural parts
being the hybrid composite considered results in the set of guided wave modes shown with solid
lines. Filled and hollow circles mark modes that did and did not interact with a 1 mm crack in the
aluminium layer, respectively. It is observed in Figure 7a that the modes of the hybrid composite are
positioned near the modes of the pure aluminium portion (marked with red dashed lines) and showed
characteristic interaction with the crack (see filled circles). Modes marked with hollow circles lay on
the modes of the pure CFRP portion and did not reflect from the crack (see red dashed-dotted lines in
Figure 7b). These modes have very small displacements in the aluminium, as shown in Figure 6b as
an example. Thus, modes of the hybrid composite show different behaviour, with either the CFRP or
the aluminium portion dominating, or both. Such an effect is very much known from the behaviour
of coupled vibrations in multi-body systems. The displacement is concentrated either in the CFRP
plies or in the aluminium layer, or within both parts depending on the mode and frequency chosen.
Such behaviour, when carefully analysed and understood, can be advantageous, allowing for damage
detection in different constituent parts of a component.

LQSODQH
RXWRISODQH
7KLFNQHVVSRVLWLRQP
7KLFNQHVVSRVLWLRQP
3RVLWLRQ 3RVLWLRQ

'HODPLQDWLRQ 'HODPLQDWLRQ
3RVLWLRQ 3RVLWLRQ

&)53 &)53

$OXLQWHUIDFH $OXLQWHUIDFH
LQSODQH

RXWRISODQH

'LVSODFHPHQWP 'LVSODFHPHQWP
(a) (b)

LQSODQH LQSODQH
RXWRISODQH RXWRISODQH
7KLFNQHVVSRVLWLRQP
7KLFNQHVVSRVLWLRQP
3RVLWLRQ 3RVLWLRQ

'HODPLQDWLRQ 'HODPLQDWLRQ
3RVLWLRQ 3RVLWLRQ

&)53 &)53

$OXLQWHUIDFH $OXLQWHUIDFH

'LVSODFHPHQWP 'LVSODFHPHQWP
(c) (d)
Figure 6. Mode shapes of: (a) Mode 2 at 475 kHz; (b) Mode 5 at 475 kHz; (c) Mode 4 at 400 kHz; and (d)
Mode 6 at 700 kHz. Red lines mark delamination placed between the second and the third ply (Position
(1)), and the third and the fourth ply (Position (2)) (counted from top to bottom).
359
Appl. Sci. 2019, 9, 4600
12,000 12,000
10,000 10,000

8000 8000
6000 6000
4000 4000
2000 2000
0 0
0 5 10 15 0 5 10 15
5
Frequency (Hz) 10 Frequency (Hz) 105
(a) (b)
Figure 7. Combined phase velocity dispersion curves of: (a) the 2 mm aluminium plate (dashed lines)
and the 6 mm aluminium-CFRP plate (solid lines); and (b) the 4 mm CFRP plate with a [90/0/90/90]
layup (dash-dotted lines) and the 6 mm aluminium-CFRP plate (solid lines). The filled and hollow
circles mark modes that did and did not interact with a 1 mm crack in aluminium, respectively. Red
dashed lines mark two aluminium modes in the aluminium, red dash-dotted lines mark two CFRP
modes. Recreated from original data from [13].
3.2. Damage in the CFRP Overwrap

Damage in the CFRP was modelled as a delamination having a 10 mm length and zero width.
The delamination was placed at two positions between different CFRP plies, as shown in Figure 3.
Position (1) corresponds to the delamination placed between the second and the third ply, whereas
Position (2) corresponds to the delamination between the third and the fourth ply (counted from top
to bottom). Modelling was performed in two separate simulation runs, and the same modes, shown
in Table 3, were used. Figure 8 shows the resulting frequency–wavenumber spectra of the propagating
modes and modes reflected from the delamination at Position (2). Excitation was performed for
Modes 2 and 5 at 475 kHz, and signals were evaluated at points positioned at the interface between
aluminium and CFRP. In contrast to the crack in the aluminium, from which only Modes 2 and 4
reflect, more modes interact with the delamination in the CFRP. These are Modes 1, 2, 3 and 5, whereas
Mode 4 did not interact with the delamination at Position (2). The results for all modes being modelled
are summarised in Table 4. Modes 4, 6 and 7 did not interact with the delamination at Position (2),
whereas only Mode 6 did not interact with the delamination at Position (1). As in the case of the crack
in the aluminium, this can be attributed to the mode shape. Every mode modelled has comparable or
bigger displacement amplitudes between the aluminium and CFRP parts (see the example shown for
some modes in Figure 6). For instance, Modes 2 and 5 reflect from the delamination even though the
in-plane displacement of Mode 2 is almost zero at both delamination positions (see Figure 6a), whereas,
for Mode 5, it is the out-of-plane displacement which is close to zero (see Figure 6b). Mode 4 interacts
with the delamination placed at Position (1), but not at Position (2). The out-of-plane displacement
is almost equal at these positions, and the in-plane displacement is even bigger in the latter case.
A contribution to this behaviour may further be the anisotropic properties of the CFRP and its layup.
Numerical modelling was also performed under 2D conditions and it could be that the guided wave
modes reflect, and scatter but, in another direction, which has not been modelled so far. In general, it
is difficult to draw fundamental conclusions at this stage as long as such multi-layered systems are not
systematically researched and described answering the question why some modes do interact with the
delamination, whereas the others do not. However, the importance of numerical modelling of guided
wave propagation and analysis, including their interaction with damage, is again highly emphasised
with the test case presented here.
360
Appl. Sci. 2019, 9, 4600
(a) (b)
(c) (d)
Figure 8. Frequency–wavenumber spectra of wave modes propagating in the aluminium-CFRP plate
and reflecting from the 10 mm delamination between the third and the fourth ply [Position (2)]: (a) for
in-plane and (b) out-of-plane components while exciting Mode 2 at 475 kHz, and (c) for in-plane and
(d) out-of-plane components while exciting Mode 5 at 475 kHz.
Table 4. Comparison of the mode-damage interaction for the damage positioned in different parts of the
structure. , modes which reflect from the damage; , modes which did not reflect from the damage.
Mode Frequency Wavelength Crack in Delamination Delamination

kHz mm Aluminium at Position (1) at Position (2)
1 100 15.6
2 475 5.3
3 950 2.9
4 400 14.6
5 475 15
6 700 7.8
7 860 11.6
4. Discussion
The example described above demonstrates that multi-layered structures are rather complex when
it comes to the description of guided waves travelling through them. To understand this, numerical
analysis tools are of an essential need. In the case considered here, the SBFEM was used to calculate
dispersion curves and mode shapes, as well as to analyse the propagation of guided waves in a
plate consisting of an isotropic metal bonded to anisotropic carbon fibre reinforced layered material.
The method allows appropriate modes to be identified and their interaction with different damage
types to be analysed.
361
Appl. Sci. 2019, 9, 4600
Results show that there are wave modes, which are sensitive to the damage defined in the
composite and metallic part of the structure, represented in Modes 1–3 with respect to the case
described here. To be therefore able to separate the damage in the aluminium liner from the damage in
the CFRP overwrap, the characteristic interaction of a mode sensitive to the damage only in one part is
needed. This has been observed for Mode 4 in the case of the crack in the aluminium liner and for
Mode 5 in the case of the delamination at both positions defined in CFRP. One has to be careful here,
in the sense that the right modes are used and that a damage in the CFRP part will not be mistaken for
a damage in the aluminium liner. This is why it is important to be able to excite modes in different
constituent parts of the component, and that chosen modes show characteristic interaction only with
the damage in one of the parts. As a remark, in the present work, a case of the CFRP being delaminated
at numerous locations across the thickness has not been considered so far, which would represent an
impact damage. Supposedly, more modes may interact with such an impact damage, and this may
result in higher amplitudes of the reflected modes than in the case of a single delamination considered
here. As for micro-cracking of the CFRP overwrap, a different approach is necessary where guided
wave modes are analysed based on the change of their phase [10] or group velocity [13], and not the
reflection from the damage.
A solution in that regard for the reception as well as the excitation of the desired modes could be
an interdigital transducer (IDT) being specifically “tuned” to the frequency and wavelength sensitive
to the respective damage to be monitored. A 15 mm pitch between its electrodes may be suitable for
the cases described here, which corresponds to Modes 4 and 5 when driven at 400 kHz and 475 kHz,
respectively. Excitation of Mode 4 will allow the damage in the aluminium part and excitation of Mode
5 the damage in the CFRP part to be identified. A design of such an IDT based monitoring system could
be to get it integrated at the aluminium-CFRP interface. The monitoring system configuration (pattern)
should be adapted to as many modes as possible being sensitive to the damage to be monitored. Here,
thin and flexible polymer materials such as PVDF can be used. Their thickness may be less than
100 μm, they can be poled to achieve piezo-electric properties, and their electrodes can be structured
in the desired manner to excite the wave mode needed [42–47]. A sketch of a proposed arrangement
with regard to structural integration is shown in Figure 9.
CFRP 15 mm Mode 5 (475 kHz)

IDT-layer
Aluminium Mode 4 (400 kHz)
Figure 9. Sketch of an arrangement based on the example considered here for an interdigital SHM
system for a composite pressure vessel monitoring.
5. Conclusions
In this contribution, the scaled boundary finite element method was used to calculate dispersion
curves and mode shapes, as well as to analyse the propagation of guided waves in a plate consisting of
an isotropic metal bonded to anisotropic carbon fibre reinforced layered material. The method allows
appropriate modes to be identified and their interaction with different damage types to be analysed.
The results show that most of the wave energy can be concentrated in a certain layer depending on the
mode used, and by that damage present in this respective layer can be detected. The results obtained
help to understand the wave propagation in such a multi-layered structure, which may further help
guided wave based techniques to be enhanced in the sense of non-destructive testing and structural
health monitoring systems to be developed for hybrid composite structures. Moreover, a concept for
an SHM system for composite pressure vessels is proposed.
Author Contributions: Conceptualisation, Y.L., J.B., C.B. and J.P.; investigation, Y.L.; software, J.B.; data curation,
Y.L.; writing—original draft preparation, Y.L.; writing—review and editing, Y.L., J.B., C.B. and J.P.; and supervision,
C.B. and J.P.
362
Appl. Sci. 2019, 9, 4600

References
1. Standard Practice for Guided Wave Testing of Above Ground Steel Pipework Using Piezoelectric Effect Transduction;
ASTM International: West Conshohocken, PA, USA, 2016.
2. Fromme, P.; Wilcox, P.D.; Lowe, M.J.S.; Cawley, P. On the development and testing of a guided ultrasonic
wave array for structural integrity monitoring. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2006,
53, 777–785. [CrossRef] [PubMed]
3. Wilcox, P.; Evans, M.; Pavlakovic, B.; Alleyne, D.; Vine, K.; Cawley, P.; Lowe, M. Guided wave testing of rail.
Insight Non-Destr. Test. Cond. Monit. 2003, 45, 413–420. [CrossRef]
4. Gao, H.; Rose, J.L. Ice detection and classification on an aircraft wing with ultrasonic shear horizontal guided
waves. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2009, 56, 334–344. [CrossRef] [PubMed]
5. Rose, J.; Zhu, W.; Zaidi, M. Ultrasonic NDT of titanium diffusion bonding with guided waves. Mater. Eval.
1998, 56, 535–539.
6. Diamanti, K.; Soutis, C. Structural health monitoring techniques for aircraft composite structures.
Prog. Aerosp. Sci. 2010, 46, 342–352. [CrossRef]
7. Mitra, M.; Gopalakrishnan, S. Guided wave based structural health monitoring: A review.
Smart Mater. Struct. 2016, 25. [CrossRef]
8. Gao, H. Ultrasonic Guided Wave Mechanics for Composite Material Structural Health Monitoring.
Ph.D. Thesis, Pennsylvania State University, State College, PA, USA, 2007.
9. Yu, X.; Fan, Z.; Castaings, M.; Biateau, C. Feature guided wave inspection of bond line defects between a
stiffener and a composite plate. NDT E Int. 2017, 89, 44–55. [CrossRef]
10. Castaings, M.; Hosten, B. Ultrasonic guided waves for health monitoring of high-pressure composite tanks.
11. Sause, M.G.; Hamstad, M.A.; Horn, S. Finite element modeling of lamb wave propagation in anisotropic
hybrid materials. Compos. Part Eng. 2013, 53, 249–257. [CrossRef]
12. Lugovtsova, Y.; Prager, J. Structural health monitoring of composite pressure vessels using guided ultrasonic
waves. Insight-Non-Destr. Test. Cond. Monit. 2018, 60, 139–144. [CrossRef]
13. Lugovtsova, Y.; Bulling, J.; Prager, J.; Boller, C. Efficient modelling of guided ultrasonic waves using the
Scaled Boundary FEM towards SHM of composite pressure vessels. In Proceedings of the 9th European
Workshop on Structural Health Monitoring, Manchester, UK, 10–13 July 2018.
14. Mair, G.W.; Scherer, F.; Duffner, E. Concept of interactive determination of safe service life for composite
cylinders by destructive tests parallel to operation. Int. J. Press. Vessel. Pip. 2014, 120–121, 36–46. [CrossRef]
15. Chou, H.; Mouritz, A.; Bannister, M.; Bunsell, A. Acoustic emission analysis of composite pressure vessels
under constant and cyclic pressure. Compos. Part Appl. Sci. Manuf. 2015, 70, 111–120. [CrossRef]
16. Degrieck, J.; Waele, W.D.; Verleysen, P. Monitoring of fibre reinforced composites with embedded optical
fibre Bragg sensors, with application to filament wound pressure vessels. NDT E Int. 2001, 34, 289–296.
[CrossRef]
17. Kang, D.; Kim, C.; Kim, C. The embedment of fiber Bragg grating sensors into filament wound pressure
tanks considering multiplexing. NDT E Int. 2006, 39, 109–116. [CrossRef]
18. Gasior,
˛ P.; Malesa, M.; Kaleta, J.; Kujawińska, M.; Malowany, K.; Rybczyński, R. Application of
complementary optical methods for strain investigation in composite high pressure vessel. Compos. Struct.
2018, 203, 718–724. [CrossRef]
19. Xiao, B.; Yang, B.; Xuan, F.Z.; Wan, Y.; Hu, C.; Jin, P.; Lei, H.; Xiang, Y.; Yang, K. In-Situ Monitoring of a
Filament Wound Pressure Vessel by the MWCNT Sensor under Hydraulic Fatigue Cycling and Pressurization.
Sensors 2019, 19, 1396. [CrossRef]
20. Saeter, E.; Lasn, K.; Nony, F.; Echtermeyer, A.T. Embedded optical fibres for monitoring pressurization and
impact of filament wound cylinders. Compos. Struct. 2019, 210, 608–617. [CrossRef]
21. Bulletti, A.; Giannelli, P.; Calzolai, M.; Capineri, L. An Integrated Acousto/Ultrasonic Structural Health
Monitoring System for Composite Pressure Vessels. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2016,
63, 864–873. [CrossRef]
363
Appl. Sci. 2019, 9, 4600
22. Yaacoubi, S.; McKeon, P.; Ke, W.; Declercq, N.; Dahmene, F. Towards an Ultrasonic Guided Wave Procedure
for Health Monitoring of Composite Vessels: Application to Hydrogen-Powered Aircraft. Materials 2017,
10, 1097. [CrossRef]
23. Yang, B.; Xiang, Y.; Xuan, F.Z.; Hu, C.; Xiao, B.; Zhou, S.; Luo, C. Damage localization in hydrogen storage
vessel by guided waves based on a real-time monitoring system. Int. J. Hydrogen Energy 2019. [CrossRef]
24. Lowe, M.J.S. Matrix techniques for modeling ultrasonic waves in multilayered media. IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 1995, 42, 525–542. [CrossRef]
25. Kausel, E. Wave propagation in anisotropic layered media. Int. J. Numer. Methods Eng. 1986, 23, 1567–1578.
[CrossRef]
26. Dong, S.B.; Nelson, R.B. On Natural Vibrations and Waves in Laminated Orthotropic Plates. J. Appl. Mech.
1972, 39, 739–745. [CrossRef]
27. Dong, S.B.; Huang, K.H. Edge Vibrations in Laminated Composite Plates. J. Appl. Mech. 1985, 52, 433–438.
[CrossRef]
28. Gravenkamp, H.; Song, C.; Prager, J. A numerical approach for the computation of dispersion relations
for plate structures using the Scaled Boundary Finite Element Method. J. Sound Vib. 2012, 331, 2543–2557.
[CrossRef]
29. Xu, K.; Ta, D.; Su, Z.; Wang, W. Transmission analysis of ultrasonic Lamb mode conversion in a plate with
partial-thickness notch. Ultrasonics 2014, 54, 395–401. [CrossRef]
30. Fellinger, P.; Marklein, R.; Langenberg, K.; Klaholz, S. Numerical modeling of elastic wave propagation and
scattering with EFIT—Elastodynamic finite integration technique. Wave Motion 1995, 21, 47–66. [CrossRef]
31. Schubert, F.; Peiffer, A.; Köhler, B.; Sanderson, T. The elastodynamic finite integration technique for waves in
cylindrical geometries. J. Acoust. Soc. Am. 1998, 104, 2604–2614. [CrossRef]
32. Leckey, C.A.; Rogge, M.D.; Miller, C.A.; Hinders, M.K. Multiple-mode Lamb wave scattering simulations
using 3D elastodynamic finite integration technique. Ultrasonics 2012, 52, 193–207. [CrossRef]
33. Luchinsky, D.G.; Hafiychuk, V.; Smelyanskiy, V.N.; Kessler, S.; Walker, J.; Miller, J.; Watson, M. Modeling
wave propagation and scattering from impact damage for structural health monitoring of composite
sandwich plates. Struct. Health Monit. 2013, 12, 296–308. [CrossRef]
34. Willberg, C.; Duczek, S.; Vivar-Perez, J.M.; Ahmad, Z.A.B. Simulation Methods for Guided Wave-Based
Structural Health Monitoring: A Review. Appl. Mech. Rev. 2015, 67, 010803. [CrossRef]
35. Gravenkamp, H.; Prager, J.; Saputra, A.A.; Song, C. The simulation of Lamb waves in a cracked plate using
the scaled boundary finite element method. J. Acoust. Soc. Am. 2012, 132, 1358–1367. [CrossRef] [PubMed]
36. Gravenkamp, H. Efficient simulation of elastic guided waves interacting with notches, adhesive joints,
delaminations and inclined edges in plate structures. Ultrasonics 2018, 82, 101–113. [CrossRef] [PubMed]
37. Bunsell, A.R. Composite pressure vessels supply an answer to transport problems. Reinf. Plast. 2006,
50, 38–41. [CrossRef]
38. Scott, A.; Clinch, M.; Hepples, W.; Kalantzis, N.; Sinclair, I.; Spearing, S.M. Advanced micromechanical
analysis of highly loaded hybrid composite structures. In Proceedings of the 17th International Conference
on Composite Materials, Edinburgh, UK, 27–31 July 2009.
39. Changliang, Z.; Mingfa, R.; Wei, Z.; Haoran, C. Delamination prediction of composite filament wound vessel
with metal liner under low velocity impact. Compos. Struct. 2006, 75, 387–392. [CrossRef]
40. Matemilola, S.A.; Stronge, W.J. Low-Speed Impact Damage in Filament-Wound CFRP Composite Pressure
Vessels. J. Press. Vessel. Technol. 1997, 119, 435–443. [CrossRef]
41. Wilcox, P.D. Lamb Wave Inspection of Large Structures Using Permanently Attached Transducers.
Ph.D. Thesis , Imperial College London, London, UK, 1998.
42. Monkhouse, R.; Wilcox, P.; Cawley, P. Flexible interdigital PVDF transducers for the generation of Lamb
waves in structures. Ultrasonics 1997, 35, 489–498. [CrossRef]
43. Ueberschlag, P. PVDF piezoelectric polymer. Sens. Rev. 2001, 21, 118–126. [CrossRef]
44. Seminara, L.; Pinna, L.; Valle, M.; Basiricò, L.; Loi, A.; Cosseddu, P.; Bonfiglio, A.; Ascia, A.; Biso, M.;
Ansaldo, A.; et al. Piezoelectric Polymer Transducer Arrays for Flexible Tactile Sensors. IEEE Sens. J. 2013,
13, 4022–4029. [CrossRef]
45. Bulletti, A.; Capineri, L. Interdigital Piezopolymer Transducers for Time of Flight Measurements with
Ultrasonic Lamb Waves on Carbon-Epoxy Composites under Pure Bending Stress. J. Sens. 2015. [CrossRef]
364
Appl. Sci. 2019, 9, 4600
46. Ren, B.; Lissenden, C.J. PVDF Multielement Lamb Wave Sensor for Structural Health Monitoring.
47. Giannelli, P.; Bulletti, A.; Capineri, L. Charge-mode interfacing of piezoelectric interdigital Lamb wave
transducers. Electron. Lett. 2016, 52, 894–896. [CrossRef]
365
MDPI
St. Alban-Anlage 66
4052 Basel
Switzerland
Tel. +41 61 683 77 34
Fax +41 61 302 89 18
www.mdpi.com
Applied Sciences Editorial Office

E-mail: applsci@mdpi.com
www.mdpi.com/journal/applsci
MDPI
St. Alban-Anlage 66
4052 Basel
Switzerland
Tel: +41 61 683 77 34
Fax: +41 61 302 89 18
www.mdpi.com ISBN 978-3-03928-299-9

Ultrasonic Guided Waves

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ultrasonic Guided Waves

Uploaded by

Copyright:

Available Formats

Ultrasonic Guided

Special Issue Editor

ISBN 978-3-03928-298-2 (Pbk)

About the Special Issue Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Christopher Hakoda, Cliff J. Lissenden and Parisa Shokouhi

Seyed Kamran Pedram, Peter Mudge and Tat-Hean Gan

Houman Nakhli Mahal, Kai Yang and Asoke K. Nandi

Houman Nakhli Mahal, Kai Yang and Asoke K. Nandi

Houman Nakhli Mahal, Kai Yang and Asoke K. Nandi

Qi Wu and Chang-Myung Lee

Christopher Hakoda and Cliff J. Lissenden

Jianjun Wu, Zhifeng Tang, Keji Yang, Shiwei Wu and Fuzai Lv

Borja Hernandez Crespo, Charles R. P. Courtney and Bhavin Engineer

Magdalena Rucka, Erwin Wojtczak and Jacek Lachowicz

Bin Wang, Yihui Da and Zhenghua Qian

Lei Zhang, Yuan Yang, Xiaoyuan Wei and Wenqing Yao

Pengfei Zhang, Zhifeng Tang, Fuzai Lv and Keji Yang

Seung Jin Chang and Seung-Il Moon

Yevgeniya Lugovtsova, Jannis Bulling, Christian Boller and Jens Prager

Received: 9 September 2019; Accepted: 10 September 2019; Published: 15 September 2019

Appl. Sci. 2019, 9, 3869; doi:10.3390/app9183869 1 www.mdpi.com/journal/applsci

Funding: This research received no external funding.

Received: 30 November 2018; Accepted: 9 January 2019; Published: 12 January 2019

Keywords: metamaterial; resonator; low-frequency; lamb wave

Appl. Sci. 2019, 9, 257; doi:10.3390/app9020257 5 www.mdpi.com/journal/applsci

(a) (b) (c)

4.1. Transmission beyond the Clamping Resonator Array

4.2. Reﬂection from the Clamping Resonator Array

4.3. Finite-Element Simulations

Figure 7. Finite-element predictions of the x-component displacement of an S0 Lamb wave at 50 kHz

Received: 11 April 2018; Accepted: 2 May 2018; Published: 5 May 2018

Appl. Sci. 2018, 8, 727; doi:10.3390/app8050727 15 www.mdpi.com/journal/applsci

2. Hybrid Ultrasonic—Fiber Optic System Concept

3. Materials and Methods

4.1. Conﬁrmation of Torsional Mode and Flaw Detection

4.2. Progressive Damage Test in SHM Mode

4.3. Individual FBG Sensor NDT

4.4. Sparse Array NDT

Received: 10 August 2018; Accepted: 29 September 2018; Published: 3 October 2018

Keywords: signal processing; SNR; split-spectrum processing; ultrasonic guided waves

Appl. Sci. 2018, 8, 1815; doi:10.3390/app8101815 29 www.mdpi.com/journal/applsci

2. Split-Spectrum Processing (SSP)

2.1. Theory of SSP

Figure 2. Split-spectrum processing (SSP) block diagram.

2.2. SSP Filter Bank Parameters

Figure 3. Filter bank parameters of SSP.

2.3. Recombination Algorithms of SSP

2.4. Implementation of the Filter Bank

fhn = fln + Bﬁlt n = 1, 2, . . . N (5)

Table 1. Recommended values for SSP parameters.

SSP Parameters Symbols Recommended Values

4.1. Experiment #1: One Saw Cut Defect

7HOHWHVW&ROODU 6DZ&XW WR&6$

P P P

Table 2. Signal-to-noise ratio (SNR) enhancement of experimental signal.

Signal 2% Defect 3% Defect

4.2. Experiment #2: Two Saw Cut Defects

7HOHWHVW&ROODU 6HFRQG6DZ&XW WR&6$

P P P

Received: 19 October 2018; Accepted: 19 November 2018; Published: 24 November 2018

Appl. Sci. 2018, 8, 2378; doi:10.3390/app8122378 47 www.mdpi.com/journal/applsci

2.1. Wave Formation

2.1.1. Pipe End

Start of Near Peak

7HOHWHVW&ROODU 6DZ&XWWR&6$

P P P

7HOHWHVW&ROODU 6HFRQG6DZ&XWWR&6$

P P P

7 ) ) )